Beyond Code: Powering the Agentic Developer Era with IBM Bob

This talk explores the transformative impact of AI on software development, particularly focusing on IBM's agentic IDE, Bob. The speaker outlines how Bob assists not only developers but also project and product managers by enhancing productivity throughout the software development lifecycle. The session highlights Bob's capabilities, such as automating code generation, identifying vulnerabilities, and providing contextual insights and explanations for existing code. The integration of AI is presented as a significant evolution in coding practices, likening its arrival to the transformative effect of the internet. The talk also delves into Bob's features, including multi-step workflows, cost transparency, and the ability to assist in legacy language modernization and overall application governance.

Before I start, I want to ask Are there people in this room who do not use an IDE, integrated development environment like VS code or Eclipse or IntelliJ or anything? Are there people who do not use? Okay, so everybody uses at some level or the other. All right, so back [clears throat] in the day when IDEs were new and you know, there are probably no IDEs about

25 years back or so, right? And at that time when you type in to all your all those new age IDEs, when you type in something, nothing would happen. You would just continue typing and you would type a lot of code to complete your work. But today developers are not writing code from scratch, right? We are most of the time staring at code that some AI has

generated or we are staring at code which is legacy, right? And and um even when you want to type something the moment you start typing, AI completes the rest of the line. Or you get you use type in a comment and AI fills in that code. So the way developers are are working is changing drastically over the you know, over the years. But then it's been very

drastic over the last few years, right? Since the advent of the LLMs and the this is this is termed as you know, one of the you know, iconic changes in the way we work since the internet. Like the internet happened something late '90s which changed completely how we work. And now with the with AI, what AI is doing is something something similar, right? We are completely changing

the way we do our work. And IBM's take on it and IBM's always been evolving. It's over a 100 year old company and it's been always evolving. And the reason it still is relevant in in in 2026 is because IBM's always been you know doing things that is best for the industry and you know adapting itself to the changing technologies. And IBM Bob is is one such

change as well. IBM's come up with this agentic IDE. It's not a plugin into VS code or something. It's an agentic IDE that helps in the entire software development lifecycle. It is not just developers. It is you know the term IDE sounds like okay, this is for developers. But at IBM there are over 60,000 employees today who have been who are using Bob and they are not

using just for code. There are there are we have program managers, project managers, product managers also using Bob to do things like estimations and planning and and analyzing diagrams and stuff like for people who are not technical they use Bob like a you know senior technical assistant who helps with technical questions. For junior developers it's a it's a it's a senior developer working with them helping them

with things. For the senior architects it is a it is a trusted source of truth that they can use to get you know opinions and answers around technical aspects. Right so with that let me get on to that. I'm not going to spend too much time on slides. I I really don't believe in doing slides for a tool like Bob. I will show most of this stuff

live. But then just a few things to set the context and the background. I already spoke about you know the the changing landscape in the in the development area. Right? The the problem really has been that AI is not new, right? It's like it's not like it just came in in 2023 or 24. It's been around for several decades now. The the way it has been showing

up is like okay, there is one AI tool for you know developer here. There is one AI tool there for QA. there is one developer one AI tool for devops and so on. But the complexity of the applications has grown so much today, right? That you know, you you can't just write a piece of code and keep it in your workspace and say your work is done.

It has to work in in a larger context in the larger application. It has to get deployed and they know you know, it has to perform in production. You have to see the developer has to see the application going end-to-end in its life cycle. And developers are not just writing code anymore, right? They are right actually creating taking requirements and you know, converting them into designs and

converting them into code and then doing QA and everything is being done by the developer himself and and then it goes out and you know, it completes the life cycle before it goes into production. So, Bob is Bob is a tool that is trying to address this requirement that that caters to these complex applications that developers need to deal with today. Right? And and what what that

means is that we need something that works with multi-step workflows. Right? Across your repositories. It's not just about, you know, your code in your in your workspace. It has to work across repositories. It has to operate across the life cycle like I talked about and and also the it has to all of this has to be done in the context of the corporate governance, right? You you

cannot break security guidelines. You cannot break security policies and guardrails when you're using AI. So, that is a very important area that AI is currently evolving on how you can do all of this stuff in a very secure way. And um And that's where Bob comes in. What Bob provides the key the these are just the key points, right? There are a bunch of others very specific

features which are that that make Bob unique, but the key features are that it's a trusted development partner. What it means is that when you generate code with Bob, it's not just like you type in in a prompt, to something and then you figure out how it uses like a lot of people have been doing that with AI in in the in the recent past that you

ask AI it will give you an answer. You you think oh my god I got the answer and you try using it and then you figure out it doesn't work. Then you go back and say oh by the way it doesn't work then the AI will tell you oh yeah you're right it actually doesn't work this is what you need to do and then you do this

back and forth for a while before you actually get the answer. With Bob all of that is is cut short and then you you you get an answer that Bob will you know deploy to your own code base in your workspace. Bob will run it it will test it it will generate tests it will run those tests make sure the tests pass if they don't pass it

will tell you why it's not passing and then it will also try to fix of course with with some suggestions from you if it's required. It will do all of that and tell you that your job is done. What what it essentially means is that it builds a level of trust you will know when you start using Bob that you know you you can trust this tool

it's not like just another tool there you ask something to the AI and it it gives you something which you don't know will work or not. Right and modernization is one big area where IBM plays strong because we we own all this my main frame and IBM i domain right. Right so there is a there are a lot of customers who still use the main frames COBOL

and IBM i which runs on AS/400 essentially. So modernizing those areas is is a is a is a very key task which which currently the LLMs are not able to do. So what we are seeing is that when you ask models to LLMs to give an answer in COBOL you will get something which most of the time is not correct. Well most people in this room would

not have done that because we not everybody uses COBOL what I'm saying that generally you know the the models are lacking in those legacy languages and IBM has premium packs for Cobol and IBM i which are used for modernization and there's also Java modernization of course for which has we have workflows I'll show it during the demo but there are Java modernization workflows which you can use

in Bob to upgrade from one version Java to the other known Liberty re-platforming and all of that stuff. All the workflows are coming built in with Bob. And cost control and transparencies is another area which is very important in the AI domain, right? We we we use AI, we get the answer, we are happy about it but then it involves some cost at the back end, right?

The it uses large language models, right? The models which are driving these assistants are called large language models and they run on something called GPU just like CPU, right? And and those are expensive. So everything that you are getting back from the AI is costing something in in in maybe in cents but then it keeps piling up very quickly, right? And then count it across the enterprise

with thousands of employees, it's going to now incur a significant cost to the organization for use of AI which means you need to use AI in a sensible way, in the right way, in a in a way which you can get correct answers in a in an effective manner. I'll talk about more about how Bob can enable that and the cost transparency what I mean what means

here is that there is you can actually see how much each session cost, right? As a developer you yourself can see what this costed you and then that the information is also propagated to the organization in a dashboard we call it as Bobalytics where you can actually see how much the AI has actually costed you. And then shift left security and features is what that means is

that you know quality and security features are shifted left in the life cycle which means that during the development itself you will do things like code reviews and in fact Bob will help you do those things code reviews and vulnerability checks. Which will improve the quality and you know the the level of security of the code. Uh we without having the code to go out first then

somebody reviews the PR and comes back and says oh there is there are 10 things you need to fix here. Then some tool will run and do vulnerability checks and come back to you again oh these 40 vulnerabilities have to be fixed. So all of that you know takes time and the productivity is immensely boosted when you use Bob because all of this is done in the

background. You can do code reviews and vulnerability checks. In fact Bob does them in the background when you are writing code. So it it reduces immensely the the time to you know take product to production. And role-based capabilities right Bob like I mentioned Bob I just touched upon some of these aspects like the Bob is not just for developers. You know there are this just shows I'm

not going through the list obviously but then you can see that from product manager to technical writer all the roles come between and there are a bunch of other roles that are not in this page but then I for every role there is something that Bob can do right. Even even if it's a DevOps engineer you can write CICD pipelines or you know log aggregation and analysis

is one thing. Our technical writers can use immensely for creating user guides and stuff. So it's not and of course the developers architects and data engineers it's a no-brainer they will use it on a day-to-day basis for a bunch of their technical queries. So there is something in Bob for all the roles and it is meant to be a tool that helps in automating the software development

life cycle. if it is straightforward and deterministic it will automate. If it is complex it will at least simplify it to an extent that then you can manage it yourselves. Right so So with that I will go ahead and open up Bob. If Have any of you been to the booth down there in the down floor, the IBM Bob booth? Okay. So, we have a booth down

there. It'll be open for the next 4 days as well. You get a sneak peek. Well, if you're attending this session, you probably are going to get most of it here. But, you you know, you can If you still have questions, you can reach out to the booth as well. Okay, I'm going to start with this application I have here. All right, let's start with the with

a simple use case where I'm going to ask to explain me a a piece of code here, and then let I'll tell you why I want to start start with this, right? When today what happens when when developers are working, right? Most of we are asking we are looking for how a specific application works, right? There is it's a issue picking If you're a developer and you

pick into your history a little bit, you will see that out of 365 days, you probably about 20% 10 or 20% of the time you're actually writing code from scratch. Most of the time, you are working with existing code, which means you need to understand how code works. You're either looking searching through your code base, or you're looking reading some documentation, or you're pinging somebody from the

team to ask on how something works. And and that's why I think this use case is important that we're starting simple here, asking Bob to explain what is the purpose of this application, explain the control flow of of one use case, right? The performance is really going to depend on how fast the internet is here, cuz I did notice that it's a little slinky since morning, but

we'll see how this goes. So, So, here is an application for the sake of the demo. I have a application called Bob was. It is It is a front end and a back end plus a database all running locally. It is It is a good example to showcase some of the features in Bob. So, here I asked Bob to explain this application, right? Just say I'm a

new developer to the team, and I want to understand how this application works. And here is Bob really going slow because the internet is not good here. All right, it it it should have got the answer by now, but let's see. So, what it's doing here is it's reading files. It's not going to read all of these files. So, it's going to smartly read some of them

which it thinks is required to answer the question. And there you go, it's already started generating the answers. And you can see it reads max 1,000 lines. It means that it reads the relevant portions of the content from this workspace, and it starts to generate a response. So, what I asked is what is the purpose of this application? And explain one key use case because then if

it starts explaining all, it's going to take time. So, I said explain one of them. So, there you go, Bob's already generating sequence diagrams and flowcharts and stuff which, you know, it's it's rare that you will have this kind of documentation in in real projects, right? We all like I mean, there's always a push to write documentation, but we know that developers will not write documentation, and

there's always a lack of documentation. And then you have to depend on somebody to ask for, you know, such explanations, or you just go through this code and try to understand. Now, if you had to generate something like this by yourself, it would have easily taken you, even if you are a super smart developer, it would not take less than, I know, half day or, you know,

a whole day to generate these kind of very detailed sequence diagrams for one use case. And you know, you can actually zoom in and see uh, more details if you're interested in. But then, you I want to bring out the the productivity boost factor here that Bob explained that this is a modern social blogging create and manage articles, social interactions, blah blah blah. It's basically a blogging

website like medium.com or something. And then it has all these different layers and technology stacks, and it's based on fast API, and it it basically brought you all the details, basic details about the application in about a few And well, not even few minutes, right? Within within seconds. And and then you also had get all the you know, the control flow charts and everything. Now, I can

go further and drill down and deep deep dive into this, but I'm not going to do that in the in the in the interest of time. I'll show you more features here. And while I am doing this, I I used the ask mode. So, the ask mode is where Bob will look at my context in the workspace, but it's not going to create anything in I mean,

it's not going to change my code. So, modes are ways Bob can be tuned to behave differently. All right? And you can basically change how Bob behaves by picking up the plan mode is more for planning out things. If you are if you have a task, you want to know how to do this, what's the best way to do this, how should I plan this out? You

use the plan mode. Code mode is where Bob will actually make changes in an agentic there's a completely native uh agentic loop implementation in Bob. So, it goes ahead and makes changes to your code. in in the code mode, advanced mode is like code mode, but it has more features like MCP servers and stuff that you want to integrate. Ask mode we talked about. Orchestrator mode is

where you know, I don't know which mode to use. I just leave it to Bob. Orchestrator mode will do the switching for yourself by itself. And then Q&A mode is something I created. Modernization mode is where uh you know, it's you know, it does things. I'll I'll talk about that later. There there is stuff around how Java modernization can be enabled with Bob in the modernization mode.

So, Q&A mode is where I said I I created for myself, so you can create your own modes. If you are a performance engineer and you want Bob to talk your own performance language, You use the uh create your own model tell Bob what you are looking for. If you are a architect, create your own mode. You know, Bob sheet that will change Bob's behavior. It will

talk in the language that you want. You can customize further how Bob behaves through agent MD files and skills and all of that stuff as well. All right, but then the basic is that there are these different modes that change Bob's behavior and you can you know tune them uh to your convenience based on what you want to do. Now, the next say now I got an

understanding of say this application. Now I'm I'm assigned a task to fix something. I I just open this file here. I can see when I open this file, I can see that there is this little icon that Bob creates. It uh it actually means that it is a there is a vulnerability that Bob has detected in this code. So, now when I hover Excuse So, when I

hover on this, I should uh I should be able to see that uh So, here you go. The Bob has identified that there is a vulnerability in this piece of code and it's offering to fix it. There are some details around some SQL string concatenation or whatever it is, right? So, and then there is an option to fix with Bob. So, I go ahead, click on that.

Bob is going to analyze this problem and see what it is and then tell me what it can do with it, right? So, again uh let me do some talking uh while Okay, this one was quicker than before. So, there is an SQL injection vulnerability, the you know, most developers would have heard of it in their lifetime sometime or the other. You always hit this problem, right?

And then Bob's uh also giving me some solution options with pros and cons. Right? Okay, so there are two solutions. And uh Yeah, now let me just go back and see what Bob did and then we'll go through the next steps. So, why this is a problem? So, it basically explained that there is a security vulnerability and why this is a problem is also being explained. What

is the severity impact is a security impact is also being explained and then the various solutions are being highlighted with pros and cons. So, now if you are developer who is seasoned enough, you know your stuff, you want to pick your answer, you can do that. Or you could just take Bob's recommendation. Bob is saying, "I'll implement you know SQL Alchemy So, I have become lazy after

starting to use these kind of tools. So, I generally pick what the AI recommends. I don't suggest that you do it all the time, but most people in our team have been doing this and one of the And that's also one of the not so nice things about AI is that blindly rely on AI to do things without understanding what the AI is doing, which is also

not so good. Right? So, as you can see here, Bob is doing it what it is doing and the reason it's asking me these questions in between whether it should save or whether it should run is because I have disabled writing files. I have disabled execution of commands and auto disabled in the sense I have disabled the auto approval. So, this is a security feature in here.

We don't want the AI to go ahead and make changes to your code randomly, right? And or run some command in your laptop and you don't everything gets gets washed off or whatever you know, right? So, these are this is a way that Bob lets you you know disable auto approval of certain things and then Bob will ask you whenever it needs to do certain things. I

generally do right and execute as disabled because those are the two areas where I I want to have control on what Bob is doing. So, in this case it's trying to run a command, so I'm going to ask Bob to just run it. Right? And I can also see that Bob made some changes to the file. It creates these checkpoints. And and these checkpoints will let us

what they Okay, I think it's trying to run some tests around here. Um so so the checkpoints are going to let show us what what changes were made by Bob and you can always go back and either leave it as is, you can view a diff of the fix that it did and you can also restore the checkpoint. If you don't like what Bob did, you can

always restore it. So this is totally transparent and that's is how I can receive it. It gives the all the trust that you that the developer needs on an AI tool because it it there is a lot of transparency on what it is doing and it will clearly that it has been done and tests have been passed and so you can see that it's also running the

tests here and confirming that everything ran fine or not. And what is it doing right now? Okay, it's actually creating a test All right. So it's doing a thorough job at it. It's not like when it said when you said there was a vulnerability, it just didn't say okay this is the vulnerability, this is what you need to do. It went ahead and gave you like three

options what you want to do. And then if you don't know I this is what I recommend you to do. You you pick that option, it actually goes ahead and makes those changes and it's also creating tests and confirming that it's all working fine. So so that's that's the you know kind of productivity boost that it brings. Now if you consider this whole use case right, this

vulnerability use case, in a typical organization if you raise a PR, it will go out and you know vulnerability tests will be done and then it will come back to you and all of that would have taken a few days already and then as a developer you would you know stare at that code for some time. You do some Googling, do some ChatGPT, find out the answer

and then do some do some fix which may or may not work. All of that noise is now reduced and you know everything is being done by the AI right you know at your own convenience and your the way you want it to be done and um And here it goes. Everything seems to be done and it's also creating the tests right now. So, while it does

that, I'll I'll talk a little more about some of the other features. So, even in the editor here, you can you can do some completions. So, Bob shows up in the chat and also it shows up in the editor. So, you're in the editor you you there is this little bulb that shows up. Uh this little bulb you you click on that and there are these three

different options. Add this to Bob will actually copy this to the chat or you can actually improve it with Bob or you know, explain with Bob. Now, now the important aspect here is that it it it's it it limits the context with which you're working. Now, this context uh is an important topic, right, in AI. AI without context is very expensive in terms of time and in

terms of money as well. If you I mean, you can use a very simple prompt and get some response and then you keep doing back and forth and back and forth until you get the right answer. But, then you would have wasted a lot of tokens, which means uh you wasted your money or your organization's money on and also it would have taken a lot more time,

right? What it would have probably taken in in what would have taken probably just 10-15 seconds would have taken like 5 to 6 minutes or even more because you didn't prompt it well. So, or you didn't provide the right context. So, in terms of providing context, Bob lets you Okay, this is taking longer than it should. Uh Bob Bob lets you you know, from the editor like

I showed, you can pick specific code blocks that you can provide as context. You can do at the rate and type in a file name and you know, you can pick that as context as well. And uh And and all of that will will limit the context that that you are you know using for the session. And it's extremely important that you provide the right context because

without that again you're you're you're just asking questions out of context and Bob will assume something and you waste a lot of tokens. Now talking about the cost transparency that I was talking about. So when I click on the prompt here, I uh these are the tokens that were used for this session and it cost around 0.55 API cost, right? API cost here is in terms of

what we call as Bob coins. So because Bob is driven by multiple models in the back end, right? And the cost is normalized and and is built in the in terms of what we call as Bob coins. Each Bob coin translates into I think um 200k tokens in terms of tokens and in terms of dollars it translates to about half a dollar. All right, so um so

this will clearly show you how much money was spent on this, right? And and then not that you will spend a little less looking at that number. It's just that it provides transparency more more to the organization, right? This number is also passed over to the organization and they can see an analytics dashboard in the enterprise version that is available and they can see clearly you know

how much money has been spent on on AI, especially you know, with on Bob what has been spent. All right, this is I think I will close this session because it's just getting trying to okay, there you go. Eat some rest finally. All right, so it it it it said it fixed all the issues and you know, it it it it showed me what all files it

changed. All these tests tests were run and everything. So this is basically what I was talking about in terms of building trust that it doesn't do a half-baked job. It will do the complete job and give you proof that it is working. Okay, so we talked about vulnerabilities, we talked about uh the modes. There are also something called uh commands that we can use. Uh when you

do a slash, you can see that it comes up with commands. Now, once your job is done, you you understood the application in the first prompt, then you did some vulnerability code fix that you were asked to do. Now, it's time to check in the code. Before you check in the code, you can do a code review. So, I type on slash review, and again, this is

a this is a shift left feature, right? We typically you raise the PR, somebody will review it, and then they will come back to you. If they like you, they will merge the PR. If they don't like you, they will express some extra defects on it and say that uh this has their their code themselves might not be complying to all the rules, but they will still

tell that your code has 10 different problems, and all of that you will have to, you know, face as part of the day-to-day development, right? So, here we'll doing a slash review is going to do a code review of your code base, and it's going to come up with findings on okay, I should have I should have run this on a different session, but anyway, Different session

in the sense I should have started a different task so it doesn't use all that vulnerability history as context, but apparently I did, and it it tried to condense the context here, which So, let me talk about the context con- densation, right? So, there's always a limit on how much these models can take in in the in the back end uh in terms of context for a

specific query. So, Bob will do an automatic condensation of the context if it's exceeding too much, and that's what it did here. Again, that involves some minor cost as well, but then if it's exceeding, it will do a condensation, which means it will summarize what's in the history, and it will take what is relevant and, you know, discard what is not relevant, and So, here it is

trying to read this code. Um this this is actually okay. It fortunately found one here. So, I Okay, it's it's doing it. It's It generally runs quite faster than this. It's just the slower connectivity here. Okay, I think it it did what it needs to do. So, here it is. It's found already some some, you know, code review comments. Let's see what to do with these. So,

the input validation, right? Very long keywords could impact performance. So, and then there it it found something like the implementation does not escape SQL like wildcards. Very typical, you know, code problems that you will see. And And sometimes, you know, even human reviewers will not be able to catch this level of detail if they are not patiently looking for things, right? So, in when it when this

is done, you can to to the location, and you can see like, "Okay, this is the problem. I want to I don't care. I just dismiss this." Or I will fix with Bob. Or I can mark as resolved. I'm not going to take any action right now, but I know I just wanted to showcase that uh you know, this this feature exists. So, So, again, um that's

how you can do code reviews, you know, security checks and stuff all in built. And all of this reduces your entire life cycle of, you know, software development heavily. Now, moving on, now we talked about a lot about developers, now, and I was bragging about Bob being useful for non-developers as let me switch to a another workspace here where I have a architecture diagram, right? So, say

I'm an architect, and I am staring at this diagram here, which looks like fairly not so simple, not too complex as well, but then not so simple. And I want to understand how this what what is this architecture all about, right? So, I'm going to ask Bob. This is an SAP, you know, business events based architecture and I'm going to ask Bob uh what is this architecture

about? All right, without even getting into the details of, you know, trying to analyze these the the connectivities and I can see that it has some six different steps to it and and all of that and uh did I reference it properly? It should have actually picked this file automatically. Okay, it did. So, as you can see, Bob can understand diagrams as well. It's not just code,

it can read diagrams, it can understand um technical details in the diagram and let us see what it comes up with here. So, it it did go through those 11 step the process here and and and you know, kind of listed it out clearly here, which is a lot more easier than to you know, see there is step one here, step two is here, three is somewhere

up here and so on. So, this kind of listed everything to me here and it also gave a basic explanation of, you know, what this architecture is. It's a business intelligence architecture for SAP and you know, a lot of stuff around it. now it doesn't stop there. Uh I I want to further ask Bob the are any issues um architecture? So, if I want to say analyze

this say this architecture has come to me for a review and I've been asked to review this architecture and see if this will work for production or is there any problem with this. Right? Typically, a lot of expertise is required to do that do that kind of review. Right? And and especially now based on the the complexity of the application and not the business domain. But so

architects today are using AI tools to do this kind of work and as you it identified Wow, it did about seven issues here. Right? It related to complexity and maintainability, performance bottlenecks, data consistency risk, scalability problems, there is also a bunch of recommendations that it's making. Now, I can pick one of these and now drill down further and ask uh now provide more details on how to

fix this problem and so on. So, the the point is that Bob can help architects as well and you know not just software architects, even performance architects. Let me show you that session here I had in my history. I should be having a session around uh So, this was one of the use cases we did as a as a proof of concept in in my IBM we

thought we should showcase this. So, here is a performance chart. Right? This is coming from a real running application which is a response time chart over uh over dates. So, the response time has been going uh moving up and down as you can see over time and uh typically a performance analyst would look at this data and now take a few hours to few days also sometimes

to why there are these peaks going up like this. What does this generally mean? Right? Uh you And without a performance analyst, you probably people will not even try to look into these kind of graphs generally. But with with a tool like Bob, we are seeing that our developers, right? Even our junior developers are able to use Bob to analyze these kind of graphs and figure out

that, you know, you don't want probably will not go ahead and start fixing these problems, but at least you don't need to, you know, wait for a performance analyst to do deep dive analysis and come back with details like the critical issues here, the extreme latency spikes, high baseline variability, and so on. So, the the point here is that Bob can also understand and numbers, right? It

It read this graph and it it figured out the dates and the numbers around it, and it clearly identified on which dates it went up and down, and and then it gave you very very very clear issues here, and it also identified know what are the potential root causes that that that are that could cause this problem. Now, and it also gave you recommended actions in short-term

what you can do, long-term what you can do, because things like performance issues are not something you can fix in a day or two, right? They It takes a while to understand where is the problem, what change needs to be done, and then you need to deploy it, test it, and then see how it goes. So, this use case I mean, this session stops here, but then

in our in the in the in our original POC, we actually went ahead and brought in the code that is corresponding to this this graph, and Bob was able to go through that code and identify clearly saying that in this code you have to go into this file, these specific lines, and then you fix these this problem, and you can see the graph will then look something

like this. It It was able to show us a graph, also the potential graphs that it will look like this if you if you fix it. There's a more that full-fledged demo is in our website on our YouTube channel. So, if you look at IBM Bob in the YouTube channel, there is a a detailed demo around this which shows this entire use case how you know performance

engineering can also be done with Bob. So that's why we named this session beyond code because Bob is not just about code that developers can write code and test code, but it's also something that you know as we saw architects can use performance architects, software architects and and and one also last thing I want to show about the other roles is how DevOps engineers can use. So

for DevOps engineers there's also a shell right? I mean not showing the shell here. You can probably see that in the booth. Um there is a terminal version as well that you that Bob will open up in the terminal and you can use it as a shell command as well. You can create DevOps scripts that will run you know CI/CD pipelines and all of that as well.

A lot of DevOps engineers have been it for in in offline mode because you they don't generally use the IDE, so they can use it as a terminal shell as well. So here is a use case which is a DevOps use case. Again, this also showcases the MCP capabilities. So there are there is an MCP server capability in Bob that you can configure MCP servers and use

them, right? So there are a bunch of MCP servers which the ones which are in green are actually in working condition. So AAP you can see a bunch of AAP servers here. AAP stands for Ansible Automation Platform. It's an IBM Red Hat platform which is used for automating you know pipelines for Ansible based pipelines which works across cloud environments, across virtual environments, physical environments and so on.

So they have Red Hat AAP has um MCP servers which we can use to do a bunch of DevOps activities and one such use case is here wherein we ran uh so the prompt you can see here is run the OpenSCAP report on this specific node and using EAP. So, we are basically running OpenSCAP as a security compliance process. So, just sitting in the IDE, and a

DevOps engineer can can run a compliance check on a remote machine, get that report and uh analyze it with Bob. And that's what this use case showcases here. So, this typically takes about, you know, 8 to 10 minutes to run this whole use case. So, not I'm just showing you the session here. Whereas, you can see Bob, you know, created a plan on how we should do

this. It then access to the MCP servers, and then it finally created a created a report here, and and and you know, with a specific date and stuff. And it said that's um that contains all the details. So, I can't think I have a copy of the report here. So, it generates a report like this, and I can then go through this report and see you know,

there are some 74 rules have failed, a bunch of things have failed, one stuff have passed. Now, this report I can then pass again to Bob. Right? And then ask, you know, okay, categorize all these failed ones, are all these failing ones related to some specific category, or do some kind of graphical visualization of these you know, issues, or you know, suggest fixes for these failed ones.

I can you know, just take it further and do more things. So, it it really reduces again the time it takes for a DevOps engineer to log in somewhere, run the report, then export that report from there, bring it here, then do the analysis, and all of that is reduced from a day or two's job to about 30 minutes to 1 hour of productivity improvement, right? So,

so that's another case, and uh before we break for questions, one last topic I want to touch upon is the Java modernization. So, there is this workspace here, which again has been run to upgrade the Java version of this specific application well, this this typically takes about 10 to 15 minutes to run. So, I wanted to just showcase a already run workspace here. I can bring out

the features of As you can see here, there is a workflow which is built into Bob, right? And it it says analyze, upgrade, and validate. So, what happens when I run this is that it will look at your workspace. It will analyze what version of Java it is you're using. And then it will ask you which version you want to upgrade to. And then it will go

ahead and once you select that, it will go do what it needs to do to do the upgrades. So, and it doesn't do all of this in a single session. So, there are a bunch of sessions that it creates. So, let me see if I can pull out some of those sessions and show the how the flow looks, right? So, So, here is when I when I

do the when I run this workflow, it's going to you know, ask like this where like you're using 1.8, do you want to upgrade to to 25 or you know, I can actually select from the drop down here. And pick a version. Now, what it does is internally, Bob comes with the recipes for these upgrades. So, there are multiple version combinations that you can do, right? So,

all those recipes are built into Bob and it comes as workflows. All you need to do is just sit back and say I want to migrate from this version to this version. Here is my application and kick play and it will do the entire migration for you and and give you a the report of what was done. So, here it did that. It did the upgrades and

and it finally it should have created Yeah, there it creates this detailed flow which shows what exactly was done, right? If you see it picked 1.8 to 25 Yeah, but 25 was picked and then it created this task for Java upgrade. And and then it it identified four or five different issues, some missing IO class, duplicate plugin, missing version ID, all typical issues that you would face

when you if you are a Java developer, you would have seen most of this, right? And then it went ahead and fixed all of those, and then it ran a verification like we saw for the vulnerability fix how it tested everything. Here also it did a very detailed verification of the whole thing, made sure everything is green. So all these reds become green, and then it creates

this final report saying, "You know, all the migrations have been upgrades have been done, and this is what you know, exactly what's been done." And all of this just costed about verification costed 0.1, and the upgrade costed 0.4. So hardly like 0.5 Bob coins is what it costed. And if you had to do this, if there is a Java [clears throat] developer here, you will know this

is not no easy task. Like doing Java upgrade is a real uh frustrating thing because a lot of things change, a lot of things will break. You fix one place, another place will break. It's And if you're sitting with a large application application especially, it is it's typically, you know, I have seen projects which run for like two or three months doing this kind of upgrades, not

on no small projects like that. I mean, we have a small example here, but then large applications will take a lot of time, and this really immensely brings down the productivity and you know, improves the productivity and brings down the time to you've seen close to like 75, 80% improvement in the um you know, performance. I With that, I will I will stop here, and let me

just make sure I covered all the things I wanted to cover. I did cover most of the stuff around the security aspects, the modes, and we did talk about explanations, we talked about vulnerabilities. We talked about um modernization, and I guess that is that is it. So I'll go back to the slides, and um I I we should be I will again urge you to strongly urge

you please go ahead and uh go ahead and uh sign up. There is a trial version, right? This This comes up trial version as well as um you know, individual plans which could be as as low as $20 or $40 or so on for individual plans. And uh for the enterprise plan or there is an enterprise plan as well that you can see the pricing on the

website bob.ibm.com. There you know, there's an enterprise plan as well if they you know, if you need larger number of licenses. All right. I will pause there. We have I think we have another 16 or probably less than 15 minutes and uh we'll pause there for any questions. Yes. Yes, yes. Skills capability I didn't showcase it but then yes, you do have skills capability. You can uh

you can actually create skills in in the in the bob folder and uh and use them. I There is actually a way to even So, you can see the skills option being enabled or disabled here which means basically it will ask you just like for the command execution and the writing capabilities, it asked you before it did. Uh even running skills, it will do that. You You

just create a dot bob folder in in here and create the skills folder inside that and it will you know, just pick the skills from there. I don't think I have a workspace here which has skills but that's how it works. So, just like you have rules.bob and /rules, you can create / uh skills as well. And one other aspect as well that I didn't um touch

upon is that you can right-click and add files to Bob ignore. So, if you have uh you know, secrets and API keys and stuff which you do not want to obviously leak out to the AI, you add those files to Bob ignore, and those files when when Bob needs to read the files, it will not read those files, making sure that none confidential data is leaked out.

Any other questions? So, does the typical sell Is conversion by 98% at least? Yes, so especially for mainframes where there is COBOL and for IBM i where there is the AS/400 RPG language, there are um premium packs that IBM is coming up with, which are not currently released, but they will be later this year [clears throat] in not even later this year in the next quarter or

so, which will handle those specific languages, but otherwise um other you open open which are not so legacy like the There are a lot of migration projects running on always, right? There's People are always migrating from something that's 10 years old to now, 20 years old to now, and so on. So, those are all you know, out of the box you can Bob for Bob to do

that. But for specifically for mainframes, there is this um um this premium pack that we are coming up with. Because the models today are they just don't understand COBOL. I don't I don't I don't I mean, they're rightly so because they have not been trained. There's not so much content to train models on COBOL and RPG and those kind of legacy languages. So, you will get some

answer when you ask today any of the models out there, you will get some answer, but a proper COBOL developer knows that that it's not the right answer. It will look very convincingly true, but it's actually not true. But there are a lot of other products being used. Sound like really expensive. Mhm. many companies are planning to go away from the back use the latest but Java.

if they like uh if they like Java maybe have use Yeah, you you use Bob so far specific modernization or migration capabilities. You use Bob to plan out the task and ask Bob to create the exact approaches, provide what approaches should be taken, and then create a plan for it, and then you know create an execution of that plan as well using using Bob itself. So, those

are typically task uh you know operations which will take uh several weeks to complete because based on the size of the application. So, Bob will help in creating the necessary scripts and necessary tools which is required to do the migration. And if you if it is feasible to use Bob itself as a chat interface, you can use or you could use it you know if it's bulk,

then you can just write scripts to as like I mentioned, it's available as a shell as well. So, you can write scripts to do the automation. More questions? So, Mhm. you come up with the product for the main frames. Like what would be the time it takes to migrate from main frame to application to application coming. A medium size application from main frame to tools. Say Java.

It really depends on the complexity of it could be I wouldn't really call out a number when I don't know it. But they are working for 1 year if it I know. Okay. application Okay. In the that way we we are I can talk to you know generally the way we have seen productivity boost, right? It's been close to we expected actually around 50 to 60% but

what we are seeing is close to 80, 85% improvement in productivity. I there are and this is not coming from some small team of 100 or 200 people using it. 60,000 IBM employees have been using it from last about 8 to 10 months now. And internally we this was released to IBM to you know use this last month but we have used this for more than a

year now. So IBM strongly believes that we don't release something that we cannot use. Like IBM is an IT company and we are a bunch of thousands of developers inside. So first we have to use it and make sure that you know be happy with it then we sell it to our customers. So that's the number that I'm done there is no exaggeration here. If you you

see you can download the trial version and you will see how quickly it can do things. There are developers in our team who have stopped writing code literally which is you know good and bad in a way but then that's the kind of productivity we are seeing. So about So for your specific mainframe I don't have a number right now but I would expect any you know

even if I I estimated based on what we are seeing right now it's it should be not less than 50 to 60%. It it could be more but at least that much. That's And and these premium packs the reason why we are releasing these premium packs is because you know if you just do it with Bobby it will not do a good job and we these premium

packs come with you know the knowledge that is required to do these migrations. So they will be much faster than you do even without that fix pack. So what does it do? It comes up with a planning and execution or else if it's somewhat actual project code it do? It does both but these premium packs the strength is that it comes with the knowledge. The main problem

today is that these large language models do not have knowledge of the legacy systems and that knowledge is built into this premium packs and that's why we need specific premium packs for for one and for RPG because you need to build that knowledge inside. And then using that knowledge and then understanding of it's be able to better understand your application. Otherwise, it's not even able to understand

the COBOL code today correctly and forget about migrating it to something else. First, it needs to understand what you have and for that it needs the knowledge and that knowledge is what comes with this premium pack. Other than this one, what other coding Other than Bob, what other coding assistants IBM uses? No, IBM does not use anything else because we when when we have Bob, we don't

have to use anything else. The models behind it are Models? Yes, the models There are multiple models driving Bob's answers. that that list is not a static list. It keeps changing, right? And we we we have a very elaborate evaluation process which runs on a daily basis because every day you see the news there's a new model which says we are better than you know what is

there. And you know, and by the time you start using it, there's already another model which says we are better. So, the industry is this this space especially is is moving so fast that we cannot just say we will use this and that's it or we The other option is like let the users pick and then they get into all this other distraction instead of focusing on

their application. They are worried about which AI tool should I use, which model should I use, how many parameters, what parameters should I set. We want to you know free the users from all that clutter and say you focus on your application, let Bob take care of what it needs to do. So, it uses multiple models in the back end and you based on your query, it

does a smart selection of the right model, which is the right model to use for this specific question and then uses that. And And if you use things like orchestrator mode, well, it's switching between different modes. Now, there are each mode will use a different model. So, it's basically you're getting the answer from multiple models and the best models out there for as of today for for

answering that Okay, did we run out of time? All right, we Let's say there are 5 minutes left and no questions. Yeah. Is that what the plan mode is about? I like it. Tell me, how does it help? So, could it be we are facing some issue with strict keyword? Okay. So, we are using correct keywords again. How does this plan mode helps? Correct. Well, the plan

mode will help you in without actually generating stuff or implementing anything, it will give you what are the different options. What is your problem? It will try to understand first. It will ask you some questions before it actually jumps in and creates a plan. So, it one has an understanding of your problem, then it will say just what are the different ways in which you can solve

this problem. And it will ask you just like that vulnerability case, in that case it was a simple problem, so it directly gave you code. But in more complex situation Yeah, complex feature, it will it will create a very comprehensive list of options and tell you pros and cons of each option. And then you can select which one you want direction you want to go. And it

doesn't end there, right? Once you select one of the options, it will then go ahead and start creating more detailed breakdown. Like, okay, you want to take this approach for this, this is what you need. Do you need all this infrastructure? You need all these many number of people and the estimated time for completion would be so much. And then, you know, just like how a project

manager or a you know, product manager would do all the planning for a for a certain product, it will provide all the details. And then, you know, it will go take it to the step that next step would be like, okay, I have the plan now, then you get How about this I mean, apparently You would not want to especially if it's a big problem. You don't

complex problem. You don't want to just pick what the AI is giving and go ahead and implement it, right? You do AI will help you in accelerating the planning process, right? Instead of you putting your head into it and and or involving, you know, five different senior people to provide inputs and doing workshops and all of that. You Bob will just do a reduce all that time

and give you a fast solution. Um I mean, it will take you to your destination in a much faster way, but then you have to still take control of what is the plan going to So, the way Bob doesn't refer to the quality of the same example. The senior most person as Okay. the number of So, that should be sufficient for Yeah, yeah, absolutely. Absolutely. And another

question, so we talked about the same picture of the nation. So, Bob was able to read >> 25 bills of problem with the year. Uh back in Yes. The example scale Yes. That's what I got. So, is that spring the picture? So, I gave it to say that go fix in the product ultimate in that case. You can you can do, but not always the problem is

that simple because for small applications, you can get all of your code in your workspace and ask Bob to fix it, but then these problems around scalability and performance and all generally they are not lying in one part of the application, right? There must be a bunch of things that are being called. Some external calls are happening. Some network delays involved. It may not always be like

you've changed this piece of code and your scalability is fixed. There could be some cases like that, but then the problem generally will be more complex than that, in which case Bob will suggest you where all you need you you can fix. And for Bob to be able to fix, you need to provide access to that code to to Bob. And not all problems um I'm one

of the person Yeah, then then obviously, yeah. I I would yeah, this this is this YouTube channel has got um that that complete demo of that response time chart which I showed, right? That that complete demo is there in the in the website there where we actually went ahead and pointed to the code and it also identified which piece of code needs to be fixed and it

also showed that if you do this fix, you will you know, the performance will be like you know, the graph instead of showing peaks like this, it will you know, flatten to a good extent. So, you can do that kind of you know, predictability analysis also with with Bob. And the last question is about the DevOps fast steps. Say like what are the uh integrated part of

this dashboard and what are the key metrics? Yes, yes. So, with that So, can we schedule something? So, like what are the DevOps in there to the one system and one screen the DevOps platform of what we are running all this? So, can Bob take some check reading and then identify some issues? Bob can help you in creating those schedules, right? I mean, Bob will Bob itself

will not run in scheduled manner, if that's what you're asking. Bob is not going to run in a scheduled manner and do something. It will start to do. So, if I what I what I'm expecting is so, monitor it and then identify if there are any issues. Like you had mentioned before, something I find it okay, there is a problem of this prime thing. If it's a

main cause. So, also it has to like what I'm asking is it has to identify some similar Uh-huh. Like I last Sorry. Say, take the example of stack that Uh-huh. Uh-huh. spring. The indexing is not proper for that particular thing. So, something like that it have identified okay, some similar issue happened earlier so, can to fix this. And so, everything from the metrics dashboard some example if

I am in the front of the dashboard I am able to see this metric. Take it. So, instead of I feed in the picture come Uh okay. it's not going to take automatically and to take from some other system also. First of all, it's it you know, it's going to involve some security risks as well. There has to be an MCP server in between that application and

Bob. And some kind of additional scripting will have to be done maybe through Bob shell you can do it either you can create a script that calls Bob and on that script you schedule through your cron job or you know some other way but you have to do the scheduling Bob doesn't run scheduled jobs. So you can create an application which will do that kind of scheduled

task and Bob will help you create that application but Bob itself will not run tasks in an scheduled way. It's not meant to be it's not designed to do that kind of task. I can take more questions offline if there are any. Thanks everyone. >> [music]