When AI Changes the Job Before the Job Description With Draup
AI at WorkNovember 18, 2026
64
00:29:5427.38 MB

When AI Changes the Job Before the Job Description With Draup

What happens when AI changes the work long before a company changes the job description?

In this episode of AI at Work, I speak with Vishnu Shankar, Chief Data Officer at Draup, about a workforce problem that is easy to miss when the public debate remains fixed on jobs being created or removed. Vishnu argues that the faster and more immediate change is happening inside existing roles. Tasks that once took days can now be completed in minutes, but people must verify outputs, understand faulty systems, recognize changing naming conventions and apply the judgment that a model lacks.

That is why Draup describes many skills as facing partial obsolescence. SQL, data modeling and similar capabilities do not suddenly become worthless. Their place within the role changes. At the same time, new AI requirements are appearing inside customer support, sales, marketing and software engineering roles. Vishnu says Draup's analysis of Fortune 500 employment data found that dedicated AI roles were growing only marginally, while demand for AI skills within existing roles rose far faster. He also describes increasing skill density, with some software engineering positions asking for 19 skills where 14 may previously have been expected.

We discuss why this creates problems for traditional reskilling. Comparing two static lists of skills misses the context in which each skill is used and ignores how the destination role is changing. Vishnu proposes starting with tasks instead. Leaders should identify what AI will automate, what it will support and what continues to require human effort. Only then can they determine which skills a person needs and whether an adjacent career route is realistic.

The conversation also turns to capabilities that rarely appear in workforce systems. Communication, for example, is too broad to assess as a single skill. In an HR role, it may mean conducting a difficult performance conversation, challenging a senior stakeholder or running a calibration meeting where managers defend their ratings. Evidence of those abilities may live in project records, manager feedback, review discussions or systems such as Jira rather than in a conventional employee profile.

For leaders, the governance question is equally important. Vishnu believes workforce transformation should be led by the CHRO, supported by the CIO's technology roadmap and informed by the CFO's budget authority. He calls for an automation register that records what was automated, where the work previously sat and what replaces it. He also argues that headcount planning becomes less informative when AI changes how much one person can produce. A capability portfolio offers a clearer view of what the organization can actually deliver.

We finish with a direct challenge to employers. Training catalogs, completion records and certificates do not prove capability. People need protected time during the working day, practice using real work with real consequences, and a credible route into a named role or family of roles. Work simulations can then test whether employees apply the skill, use the available tools and catch planted errors. Vishnu also explains how unrealistic experience requirements, overloaded job descriptions, slow hiring processes and weak compensation can all be mislabeled as talent shortages.

Is your organization measuring the number of people who completed AI training, or whether those skills are appearing in the work they produce? Share your thoughts and experiences with me.

Useful Links

[00:00:00] [SPEAKER_01] Once a year, I open up a handful of places for people who are serious about launching their own podcast, but have never actually gotten around to it. And over the last 11 years, I've helped entrepreneurs, business leaders and Fortune 500 companies all launch their own shows. And here is how it works. You record your conversation on Zoom, Teams or whatever works best for you.

[00:00:23] [SPEAKER_01] You send me the audio, an episode title and short description, and I will handle everything else from setting up the podcast host and distribution to getting your show on Apple Podcasts, Spotify and the many, many other podcast platforms out there. I will commit to working alongside you to help you launch your first season of 12 episodes in the bag. So you can start 2027 by ticking a season of podcasts off your bucket list.

[00:00:52] [SPEAKER_01] Think of me as your not just personal podcast producer, but also your accountability partner to make sure that you actually deliver those 12 episodes into podcast land. Now, I only offer this once a year with limited places available. So if you want 2027 to be the year you finally launch that podcast, head over to techtalksnetwork.com. Send me a message, send me a DM or even an audio message. I'd love to help you make it happen.

[00:01:23] [SPEAKER_01] What happens when AI changes the work months before anyone even updates the job description? The gap is already creating confusion for employees, hiring teams and the leaders that are expected to turn an AI investment into a workforce plan.

[00:01:42] [SPEAKER_01] Well, my guest today is the chief data officer at DRAUP and he explains why many skills are becoming partially obsolete, but without disappearing. And we'll also discuss why existing roles are absorbing new AI requirements and why counting course completion tells us very little about whether someone can actually perform the work.

[00:02:07] [SPEAKER_01] So we're going to discuss today the shared responsibility of the CIO, CHRO, CFO, way too many acronyms there. And the difference between a genuine talent shortage and an unrealistic job description. And for the rest of us, why employees need protected time, real practice and credible destination for re-skilling.

[00:02:31] [SPEAKER_01] So if your AI plan and people plan are running on different clocks, this is a conversation for you. Well, enough for me. Let me introduce you to my guest now. So thank you for joining me on the podcast today. Can you tell everyone listening a little about who you are and what you do?

[00:02:51] [SPEAKER_00] Yes. Thank you for having me here, Neil. I'm Vishnu Shankar. I work with DRAUP as the chief data officer. The simplest way to put what we do at DRAUP is we are a workforce transformation platform. And what I mean by that is we help CHROs, chief AI officers help convert AI disruption into a workforce plan that they own.

[00:03:17] [SPEAKER_00] And what that means is, again, how would you quantify AI's impact on the company level down to the function, down to the role? How would you model ROI around those decisions? How would you build a skill architecture around the work that remains? And with all of this, how would you support the leader with evidence rather than a general theory? Right. So that's kind of what we do, what roles to redesign, what skills to build and what the payoff actually is, to summarize.

[00:03:46] [SPEAKER_01] Thank you for joining me today. And one of the things that put you on my radar, you set off my tech spidey sensors when I read that you'd suggested that a large share of IT skills face partial obsolescence by 2027. So tell me a bit more about what does this mean in practical terms for a person doing their job in IT today? What were the big takeaways here?

[00:04:11] [SPEAKER_00] Right. That's quite a dramatic statement, isn't it? Yeah. Imagine this, right? And picture an analyst three years back working in a supply chain team. A large part of her work would have been to, say, create a demand report. And what this means is she'd pull data out of three different systems, clean it up and produce the data, model the data and create a report finally. Right.

[00:04:38] [SPEAKER_00] And this would probably take three days of her work. And it was the hardest part of her work. Right. That was genuinely expected out of when she got hired or when she got promoted. But today, things have changed. She describes what she wants to model. Everything gets done in a few minutes. She has a working draft. But then her job description has still not changed. It still says SQL. It still says data modeling. Nothing has been removed from it.

[00:05:06] [SPEAKER_00] But they've changed their position. Right. They're not as important as they were two years ago. And what's expected out of her is it's not has not gone away. She still needs to verify the report that the model generated. She still needs to say, hey, what's working right, what's not working right. She's the only person who knows that one of the systems may not be working properly or it returns data a little late.

[00:05:34] [SPEAKER_00] Or the October numbers are usually spiked. Or are there something that changed in the naming convention? And she's the only person who can say this forecast is wrong. So what's expected out of her has completely changed. Her JD has not. The skills still remain there, which is why we also call it partial absolutes. The importance of those skills have reduced, but they've not gone away completely. And then it's also very hard for organizations to see this because the whole work has changed underneath,

[00:06:02] [SPEAKER_00] while the job description or how she describes her work has not completely changed. So that's kind of what this would mean. Yeah. So let me know if that answers your question.

[00:06:15] [SPEAKER_01] No, it really does. Because when I asked the question, as you said, it sounded quite dramatic, but it's quite the opposite, really. Because another thing that you argue is that AI is changing the task mix. Do that again? You argue that AI is changing the task mix inside roles faster than it's eliminating entire roles. So there's a lot of positivity here and a lot of opportunities and optimism.

[00:06:40] [SPEAKER_01] So what technology jobs are you seeing that distinction most clearly where the role is changing rather than being eliminated?

[00:06:48] [SPEAKER_00] And this is something I have seen empirical evidence for, right? Yeah. This particular transformation that's happening where roles are changing faster, the task mix is changing faster than role elimination, which was what was originally anticipated with the whole AI wave.

[00:07:06] [SPEAKER_00] In fact, when we analyzed data of the Fortune 500 companies and their jobs, we would have expected typically that AI was eliminating roles that, you know, we typically thought will be automated largely. But that was not the case. Most of these roles remain flat. If you look at overall trends and none of the roles were disappearing. And you would also expect the AI roles to grow up, which also was not the case. They were also not growing so steeply.

[00:07:35] [SPEAKER_00] Dedicated AI roles, the reality was dedicated AI roles were only growing marginally. And what grew was underneath, right? Which is AI skill requirements inside the roles that already existed. For example, if you look at customer support, nearly 25% of the, you know, there was a 25% increase in requirement for AI skills and customer support, right?

[00:08:02] [SPEAKER_00] Sales and marketing also had very high increase of AI skills needed. So typically, depending upon the function, anywhere between 16 to 25% had really high AI skill growth. And there's another angle we should look at, which is skill density. If you look at typically the number of skills required from role, it used to range around about 15 skills a role on an average or less.

[00:08:27] [SPEAKER_00] But what's happened now is something like a software engineering role needs 19 skills today, right? Whereas it used to be 14 skills. So number of skills required per role has also increased, what we call a skill density, right? And, you know, this is true for even, and then I think one more thing to add is how skills are changing so fast. Skills that are needed from the roles are changing so fast. For example, if you, this is true even for an AI builder role.

[00:08:55] [SPEAKER_00] If you look at AI roles, the roles that are actually building AI models, even for them, the skills are changing so fast. Between last year and this year, there are new skills that have come up and increased in demand so much. For example, agentic orchestration or rag models or evaluation and guardrails for LLM models and so on. Whereas certain skills like scripted chatbots or feature engineering or manual data labeling are all falling away, right?

[00:09:24] [SPEAKER_00] So these roles are not being relabeled. They're just changing underneath. So this is what we mean by the underlying roles changing so fast, the tasks expected of them changing so fast, and not roles being eliminated themselves.

[00:09:42] [SPEAKER_01] And for any leader that's listening and they're looking at AI in the workplace and how it's impacting their employees and their teams, how can they help their company identify adjacent skills that make re-skilling much more realistic rather than just reacting to AI forecasts with broad layoffs or just generic training and all that bad stuff that we see out there and they want to do something different. Anything that you'd advise here?

[00:10:09] [SPEAKER_00] So I think re-skilling itself, where it needs to be arrived at, how it needs to be arrived at, I think it needs to follow a certain framework that we've been talking about, which is the build by borrow bots framework. And I think when you're looking at any kind of workforce transformation initiative, it is these four layers. But within the build layer is where the re-skilling as an outcome lives, right?

[00:10:34] [SPEAKER_00] And even for re-skilling, the typical approach has been, I have a skill list on this side, I have another skill list on this side, depending upon the intersection or the non-intersection, for lack of better words. You decide, hey, what is the most feasible path to take and what are the skills to learn? What is the skill gap? But the problem with that, again, is you're comparing today's role versus today's version of another role.

[00:11:02] [SPEAKER_00] And you have two different skill lists for those roles. Whereas skills itself changes, the meaning of the skill itself changes depending upon the context of where it's being done, right? Like I may use a Python for software development, I may use a Python for an algorithm development, I may use Python for software testing, for example, right? So same skill can be used for three different purposes.

[00:11:27] [SPEAKER_00] So without the context of what that skill is needed for, preparing this skill list and identifying adjacent skills for re-skilling does not make sense. So the ideal approach would be to look at what is the role of this person today and what is the future version of this role look like that is needed for the organization. And then look at what is the task performed by this role today?

[00:11:49] [SPEAKER_00] What is the future version of role looking like and what are the tasks they will do, which would be tasks that are eliminated after the things that AI would automate, things that AI would augment, the things that would genuinely require human effort. And then based on those tasks, identify what skills are required and then plan the re-skilling journey, right? So that's typically how we would recommend approach.

[00:12:14] [SPEAKER_00] One should approach re-skilling in general and not look at static task skill lists. Yeah.

[00:12:21] [SPEAKER_01] And of course, workforce data that can map qualifications and job histories and all those traditional metrics. But it often misses things like judgment, communication, critical thinking, institutional knowledge, all the things that are crucial now we're working with AI. So how should leaders better account for those maybe less visible capabilities that are really required when working with AI?

[00:12:45] [SPEAKER_00] So this is one of the things that data that's available today, at least publicly, has a honest limit, right? So there's none of the data sets, labor market data sets that are available today can say whether a person has a judgment skill. It's not written down anywhere. And we also tend to generalize some of these skills. There's a naming problem with these skills. For example, we say communication. It's not a capability. It's a category of skills, right?

[00:13:14] [SPEAKER_00] And you can't assess the category of skills. I'll give an example of why I say that's a category. For example, if you are analyzing a HR role, right? And HR generalist, for example, or HR business partner, what does communication mean? Or what are those skills that's required from a HR within the context of communication as a category? Would they be able to have a courageous conversation with a senior stakeholder?

[00:13:41] [SPEAKER_00] Would they be having a difficult conversation with an employee who's not performing well? Or facilitating a calibration session where managers have to defend their ratings for their employees? These are nuances that's actually expected out of an HR, but doesn't actually get tracked or quantified or measured. So that's where a lot of this gets done. Now, where do we look for this data?

[00:14:11] [SPEAKER_00] We typically look for this data in the HR is. It doesn't exist there. We look for this data in job postings, job descriptions. It's not actually well-defined there. Where some of this exists is project records or any mentions. It could be your JIRA systems where people log all their work. It could be your review discussions and your documented work of what your manager feedback has been.

[00:14:35] [SPEAKER_00] So I think some of these data sets is where a lot of this knowledge lives around employees' capabilities, employees' judgment skills, or the actual human-driven skills, right? And I think that some of this is recoverable from evidence that exists. Some of it may not be. It probably resides in the manager's heads. And the goal is to see how much of this we can leverage to actually understand our workforce better.

[00:15:00] [SPEAKER_01] Do you need AI agents that you can trust? Well, with an AI data layer providing real-time connection within your data platforms, you can trust your agents to provide accurate solutions. So scale your business by trusting your agentic AI, accurately getting the work done for you. Trust its capabilities with Denodo. And you can do that by simply visiting denodo.com to learn more.

[00:15:28] [SPEAKER_01] And again, for people listening, what should a shared workforce plan between, let's say, the CIO and the CHRO contain? And I appreciate this is a big question, but which decision should neither function make on their own?

[00:15:44] [SPEAKER_00] Here, I think there's also usually a third stakeholder who comes in, the CFO, who has the budgets. Yeah. But then between the CIO and the CHRO, with respect to the workforce transformation, I think the owner who holds the pen, it should be a CHRO-led problem. It should be driven by the CHRO. On the surface, it looks like a technology transition.

[00:16:06] [SPEAKER_00] But I think the CHRO is the only person who's being accountable in the building for the capability that the company has three years from now. So I think it's a decision that the CHRO has to make, but they cannot make it a law because most of what it requires also sits with the CIO or the engineering leadership.

[00:16:27] [SPEAKER_00] Because it also presents a tooling choice, which is evaluated on cost, cycle time, error rate, and several other factors where the CHRO needs the help of the CIO. The consequence still lands with HR, but the plumbing, for lack of better words, is enabled by the CIO. And these two things can't run on two different clocks, right?

[00:16:49] [SPEAKER_00] Because the CIO says he's done anything, he's implemented all the tools, he thinks everything is automated, but nothing shows up this quarter, nothing shows up next quarter. The problem amplifies four years later, and then the CHRO is the one being accountable, right? I think the shared plan needs two things. First, an automation register with a workforce consequence attached to each hire. What did we automate? Which carrier step did that work sit on? And what replaces the step?

[00:17:17] [SPEAKER_00] It's a short document not being written today. Second, shift from planning headcount to planning capability. When AI changes what one person can produce, the headcount number stops carrying the information, right? What you actually need is a view of the capability portfolio that requires the CIO's roadmap as an input and the CHRO's ownership of the outcome. So a lot of this can stay separate, but the tool selection, for example, the tool selection is a CIO call.

[00:17:47] [SPEAKER_00] Compensation structure is a CHRO call. The middle layer needs both, right? So that's how this has to be balanced today.

[00:17:55] [SPEAKER_01] And for a technology leader that is currently planning the next 12 months with one eye on 2027, what would you say is the most useful way to inventory skills, select training, and then measure whether people can apply what they've learned? What kind of metrics do you put in place there?

[00:18:15] [SPEAKER_00] So I think three things you're asking for, right? One, inventory. Two, training. And three, machine. Inventory, I would say we should start looking at your own data. One, your own job postings and work backwards from there, right? So you have your own job postings. You look at what you're asking for from the market for the last 12 months. And you compare it with what your people actually demonstrably do, right?

[00:18:42] [SPEAKER_00] So that gives you a very clear picture of what are you looking for in the market versus what are your people needing or what are your people doing? And this divergence or convergence between this itself tells you what's the inventory that you need, right? Second, in terms of training, the most common failure is you give a catalog of 300 training programs to your employees. Having a conversion rate in the low teens and tells you nothing when you look at the data.

[00:19:11] [SPEAKER_00] It tells you just people are probably busy. They're not even taking these trainings. And, you know, with what we discussed earlier about how skills are becoming obsolete really quick or partially obsolete really quick, what organizations actually need is not a curriculum but learning velocity, right? How fast can this organization move when the requirement keeps changing every three months, every six months? Because it is already changing. It will change further.

[00:19:38] [SPEAKER_00] So learning velocity of your organization becomes extremely critical. And that's what is going to drive training. Measurement. Again, measurement is largely today looking at completion or certain number of certifications. Completion is not capability. Certification is not capability. The only evidence that counts is the skill showing up in work that's done, right?

[00:20:04] [SPEAKER_00] Like you ship some work and that is the only true evidence, only true outcome you can measure. And it is harder to measure, but it is the only thing worth measuring, right? If, for example, what we've been proposing about how people should hire as we have this 90-day version of how we should redesign hiring, the same instrument would work for measuring internal capability, right? You look at, you give people, put people in a work simulation rather than an assessment.

[00:20:34] [SPEAKER_00] Give them real work to do. Give them real messy data. Give them tools that they can genuinely use, that they can augment in their work. Plant an error. Test them out whether they're catching it. Then score them across all of these dimensions rather than just saying, you know, did they clear it or not clear it, right? So that's what is something that is measurable and would add value than a certificate or a completion record.

[00:20:58] [SPEAKER_01] And I think it's easy to see why many workers are fearful of change or concerned about their future, because so many employees, they're often told to reskill, which is great. But on the flip side of that, without being given time, practice, or a credible route into a new role. So what responsibility should employers accept here if they want workforce planning to really work and improve morale across teams as well? And just ensure that nobody gets left behind.

[00:21:28] [SPEAKER_00] Yeah, that's again a very common mistake that organizations tend to do. Reskilling is mandated, but now if the managers give the protected time for that reskilling, people end up, you know, doing it namesake, they do it over evenings, do it over weekends just for the sake of completing it. The employee absorbs the cost, but nothing leads anywhere, right? So three things that employees, I believe, should do if they want this to work. One, absolutely time.

[00:21:58] [SPEAKER_00] And this would be inside of actual working day hours. So it should not be clawed back by delivery pressure. Number two, practice on real world work rather than exercises as part of the curriculum or learning. Because the skill is only durable if it's applied on something which has a consequence, which is actually has something at stake. Third, a credible route towards which that practice is happening.

[00:22:27] [SPEAKER_00] For example, a named role or a class of roles that the person is designated for or a path that is leading them to. Without the third one, especially the first two amount of professional development, it has value, but it's still not workforce planning, right? Because employees are smart. They can tell the difference immediately if this is a serious initiative.

[00:22:51] [SPEAKER_00] You know, if the program has a destination or it's just a random reskilling program without an outcome at the end, right? So without, I think, these three things, reskilling programs do not make sense. And the encouraging thing is some companies are already doing a version of this properly. We've seen companies like Accenture put real money into this.

[00:23:14] [SPEAKER_00] They even reported recently that, you know, they took their data on AI population from 40,000 to 77,000 employees in two years. They also kept hiring entry-level people, right? While changing what the entry-level work means. This is a good example of a serious commitment to reskilling programs and a useful benchmark for what this thing costs when you mean it.

[00:23:39] [SPEAKER_01] And finally, again, to give anyone listening a valuable takeaway here, especially leaders, if they're looking to distinguish a genuine skill shortage from a hiring process that is asking for unrealistic combinations of experience and qualifications and emerging AI capability, any tips or advice or anything you would add to help leaders distinguish between the two?

[00:24:04] [SPEAKER_00] The thing is, skill shortages are real today, right? For example, we did a study where we found that the global pool of genuinely AI-capable workers stands at about 2.2 million workers, with US having about 300K or so. And just for the Fortune 500 demand, that's not enough, right?

[00:24:28] [SPEAKER_00] But then, many times what gets reported upward as skill shortage is not truly that. While this problem is real and true, and it's not going away quickly, quite often what seems to be going wrong is a few things. For example, one of them is what's proper arithmetic sometimes. It looks very simple, but happens, right? You would see job postings which are requiring years of experience that are more than what

[00:24:56] [SPEAKER_00] sometimes the technology has actually existed. It may ask for five years of something, whereas the market would really not have anybody fitting that requirement. And it's mostly some lazy job description creation happening from an older template and so on. Second is combinations. Like we discussed earlier, skill density is becoming high. What's expected out of one role is becoming high.

[00:25:24] [SPEAKER_00] So postings are asking for three or four distinct skills than they did a year ago, right? And some of these are genuine role enrichment, but some of these are budget constraints or headcount constraints being converted into single JD, right? So they're limiting, trying to fit in as much as possible into a JD. And this also sometimes ends up becoming projected as a skill shortage because those combinations don't exist that easily in the market.

[00:25:54] [SPEAKER_00] Third is more of a time to fill problem. A time to fill, for example, if, and there are two ways this can go, right? One, a role may sit open for months, but once you extend an offer, it gets accepted. So in this case, the market has the people, but the process has an issue. The other case can be you're finding people, but offers get rejected. This is usually a compensation problem.

[00:26:22] [SPEAKER_00] Neither of them is a skill shortage, but sometimes gets reported as one, right? So genuine skill shortage usually would be where the capability is new. People who have it are concentrated in a very handful of companies, a very limited set of companies. Compensation for that is being priced really fast. Like today you're pricing it somewhere, two months later, the same skill is being priced much higher because the demand is so much. And not all of this gets landed as skill shortage sometimes.

[00:26:51] [SPEAKER_00] So that's, I think that's, that's what I would say my read on this.

[00:26:56] [SPEAKER_01] Absolutely love that. And we've covered a lot in a short amount of time today and also given people listening quite a few valuable takeaways and I'd love for them to feedback and let them share their stories and experiences with us. But for them that want to find out more information about you and everything we've discussed today, where would you like to point them?

[00:27:16] [SPEAKER_00] You can definitely check us out at our website, draup.com. A lot of the research that we do, a lot of the papers that we publish, blogs we write are there and we're quite active on LinkedIn. I'm very active. My other leaders are very active. Our page is very active. You can also check us out on LinkedIn. We do a lot of work there too.

[00:27:34] [SPEAKER_01] Awesome. Well, I'll include links to everything that you mentioned. There's so many big changes happening at the moment, but so much room for optimism and excitement for a lot of opportunities are coming, but there's just a responsibility on all sides to better invest in the skills that we have. But thank you so much for shining a light on this. Really appreciate your time today.

[00:27:54] [SPEAKER_00] Thank you so much, Neil, for having me here. I look forward to when this podcast comes out. Thank you.

[00:28:01] [SPEAKER_01] I think one of the many things that stayed with me from this conversation is that workforce planning cannot stop at headcount, course catalogues or a list of the latest fashionable AI skills. It's the work itself that has to be examined. Which tasks will disappear? Which will be augmented? Which will demand human judgment and critical thinking? And what evidence proves that a person can perform them?

[00:28:28] [SPEAKER_01] My guest offered employees a very fair challenge today. If reskilling happens during evenings and weekends with no protected time and no named role at the end, if that is the route that you take, the employee carries the costs and the business learns very little. So again, thanks to my guest for joining me today and giving leaders a very refreshing way to connect technology decisions with capability, career paths and measurable work.

[00:28:58] [SPEAKER_01] I'll include links to everything that we mentioned there, including the workforce research. But over to you. Are your AI plans changing job descriptions, training time and career routes at the same speed as technology? What are you doing about that? Remember, you can get me at techtalksnetwork.com. As the summer comes to an end, I'm taking the show back on the road. I'll be attending a lot of tech conferences until the end of the year.

[00:29:26] [SPEAKER_01] If you are at any of them, please drop me a line. Let me know. If not, send me an audio message, connect me on LinkedIn or have a look at the 4,000 other interviews we've got. But that's it for today. So thanks for listening as always. Speak to you tomorrow. Bye for now.