Future of Technology Consulting in the GenAI world


As always, these are just my own personal opinions

There isn’t a consulting firm out there today that doesn’t have AI and specifically GenAI included in their respective stories. Over the last year or two – consultants and system integrators have done a huge lot of proof of concept work. I don’t know a single client who doesn’t have a clear mandate to derive value from GenAI. This mandate is usually from the board but at a minimum it’s from the C suite. Analysts have been trumpeting GenAI as well.

Having been in this industry for more than a quarter century now, I have seen versions of this play out from ERP in the 90s to web to mobile to cloud and now AI.

So back to GenAI – why is it that everything looks conducive to explosive value add and yet no one seems to be putting massive transformative projects into production?

There are a few common themes

  1. People have only a vague idea of GenAI and hence most people are looking for use cases that scale with GenAI
  2. Lot of idea generation sessions have happened but there is no framework to decide which ones to bet on
  3. For those ideas that seemed promising and hence piloted – the first million dollars of value was easy but that didn’t seem to translate to a scale of tens of millions.
  4. The quality of data makes it quite hard to translate the value seen on PPT to hard dollars that show up in the general ledger
  5. Unit economics don’t seem to work in favor at massive scale and business cases don’t hold up. CFOs are especially worried given they already went through one nightmare with unpredictable spend on public clouds.
  6. Last but not the least – regulatory frameworks have been a challenge for many use cases and deployment models

What we keep forgetting is the simple fact that GenAI is still quite a nascent technology. The good thing about it is that we are past the point where we need to worry about whether it is useful or not. Now the challenge is largely operational in nature – where we need more engineering than science, and more product management than marketing

The world of POC and pilots does not really look all that transformative in my opinion. Sure it’s incrementally better and has some productivity gains for sure. As an engineer, I love how GenAI helps me with code completion, test generation and so on. I enjoy the geekiness of an LLM helping me with emails. Since I am a decent engineer and since I believe I can communicate quite effectively myself – I won’t miss that help if I am told that I can’t use it from tomorrow. These are all just good things to have and gives me confidence to think about the bolder transformations that will follow

Where the small productivity improvements help a lot is that it will conserve time and cost to invest into the bigger things that come next. So I do think that what the tech world has achieved so far is quite useful.

I think the strategy consultants will need to learn GenAI in some serious depth to evolve the frameworks used to qualify which use cases to invest in. Jargon – as much as our industry loves it – ain’t gonna cut it. They will need a much better grounding on the financial modeling of a GenAI deployment as well when the world moves from one model working monolithic to many models working together in a compound system

I have always been a big fan of open source. I think the tech consulting world will largely make use of open source over the long term.

A lot of conversation on GenAI is centered around the idea of a model. Model is absolutely the foundational building block – but I don’t see massive deployments in the enterprise landscape based on one model, and nor do I see GenAI as a stand alone technology.

A minimally sophisticated use case today needs a model and usually some RAG construct to go with it. The future though belongs to compound systems – many different highly specialized smaller models working together in some well orchestrated manner and using both GenAI and other technologies. A crude analogy from the past would be ERP and CRM which were touted to be the one answer to all questions – and when we look back we can see that most of the work happened in integrating them with a lot of other things in the enterprise landscape.

This needs a lot of upskilling for the current tech talent – and will need some serious interdisciplinary training. It will need more training in systems thinking than the consulting world historically is used to and that won’t be an easy transition for many. And given the speed at which AI develops – the tech community will have to spend countless hours learning to just keep up. That’s again not something that we have seen at scale. I am not even sure if the HR teams around the world are capable of handling that scenario, not to mention the CFO teams sweating about modeling the investments and the return on capital. “How we work” will have as much disruption in our organizations as the tech disruption will be in the engineering world.

Let’s consider something like optimzing back office process – say accounts payable – as an example on how the “way we work” will change. Traditionally we would use some lean six sigma analysis to optimize the process, use some tech to automate what we can with OCR and some elementary AI , and then depend on labor arbitrage via BPO to save most of the cost. A lot of companies have taken out hundreds of millions of dollars already this way – there is only so much left to squeeze out with the traditional approach.

If we look at an AI first approach – then the BPO staff would need to learn to properly label the data they are working on . Then an AI team will need to use that to do some supervised learning to get a model trained. To some degree – the better consulting firms all do some version of this already. But we do know that learning by mimicking only gets us to a certain level of efficiency. So we will need to do even more AI work – where the model learns via trial and error (like reinforcement learning). From that point the team will need to build a compound system to orchestrate the piece parts. I am sure those of you who are from the consulting world can already extrapolate the changes in operating model a consulting business will need to pull this off.

As the AI agents become more mainstream and start working with humans – perhaps even interchangeably- there are a host of other aspects to consider. We will need a modern version of the current HCM suites to onboard, train, performance manage and retire the digital agents. It will need a whole new set of integrations with finance and costing apps. And all this assumes that the governments all around the world will get smarter with appropriate regulations

What about the skills we already have? Does anything at all that we know now survive this massive shift?

I do think that despite the need to upskill constantly – many of our existing skills will transfer over just fine. As an example – let’s consider data management. There is no way that any of this AI goodness will happen without great data management ( quality, governance, security, lineage and all that). If anything I think GenAI will make data management the new black. GenAI will make it easier to execute data management for sure – everything from discovery to code fixes will be much easier but the core principles will all be still transferable

I am a massive optimist when it comes to technology. I am perhaps a little too excited about the fun challenges I will get to solve in the next ten years. If I have any regret – it’s just that I am not an engineer in my twenties anymore. I guess my generation had our share of fun with other technology shifts and still will manage to play a small role in this one

I think Satya Nadella might be wrong about the death of SaaS


I will start with the usual – what I post here is strictly my personal opinions. It has nothing to do with my present or past employers

I watched this fascinating interview with Satya Nadella where he predicted SaaS will be dead and Agentic AI will take over.

I am a huge fan of Satya – and I listen to him closely every time I get a chance. He is one of the most grounded technologists of our generation. But this time – I felt his prediction of the death of SaaS is exaggerated

I do think that AI agents are the next generation of innovation in software and it will be disruptive to how SaaS has historically worked. But if anything I think the actual impact to the SaaS business is that it will grow massively in near future itself

Let me explain my thinking and maybe someone who reads this can correct me

The big disruption essentially is the co-existence of human labor and digital labor in future. As digital labor – or AI agents – become more common place, it will become a necessity for the current HCM SaaS apps to put them into scope. Think about it – those agents are going to do types of work that humans have done in the past – and they probably will be a mix of owned and rented entities, much like employees and contract labor today. Their work needs training, performance management, security and so on. I would expect workday, oracle, sap etc to evolve to cater to that reality in short order

Or think about trading partners for a business like customers and vendors. A simple example would be me as a consumer offloading my grocery shopping to an AI agent (sat next generation of Alexa). If the agent is the one in charge – there is no reason for the vendor to market to me via emails and snail mail and phone calls – they need an agent on their side to pitch to my agent through a digital protocol. Think about the changes needed in Adobe and Marketo and so on to cater to that new world.

So just from added complexity of the scope itself – SaaS would grow, not die

Now let me poke at the other assertion by Satya that the business logic will be taken over by the multi repo agents, and then the backend systems collapsing

If you look at how the major SaaS apps treat common “objects” like customer, purchase order etc – you will notice that there is literally no standardization across the vendors today. That is why integrating SaaS systems is a big cost and a painful thing in most client landscapes even when vendors claim their API’s allow seamless integration. On top of that – every client has nuances built via additional configurations and customizations. So even if every vendor agrees to use their metadata and data to train an AI model – it still takes a lot of effort for an AI agent to understand the system that is in place for a given customer

The problem gets worse when the process logic is not held in all in SaaS but it an external system like an RPA or BPM system – which is a very common pattern in enterprise context. In most cases – these are not even documented anywhere and is just in the brains of a few people. There are discovery tools that use multiple techniques to analyze the landscape to make sense of such information – but so far I haven’t seen even 50% of a complex landscape being auto discovered without massive human intervention

Business logic for most applications is quite deterministic and stateful – for good reasons like the need for consistency, auditability, security, compliance, performance and so on. There is barely any room in that set up for even a tiny amount of hallucinations. Even rock solid deterministic systems still need a lot of work and have errors that need human effort to reconcile and fix.

Satya abstracted SaaS to CRUD operations on top of a database. That’s not wrong – but obviously it’s not that simple either. When you pay an employee their salary – it is not just the HR system that needs to be updated, it is also the ledger. Not all those integrations are clean interfaces and many are not even externally exposed. So the ability of an Agent to function in the multi repo way that Satya explains it is not an easy one to pull off in practice

While I don’t agree with Satya that SaaS is on its deathbed for all the reasons above – I absolutely think Agentic AI will be a great addition to it and will add tremendous value. Since there is already plenty of hype on what all Agentic AI can do – I don’t think it’s useful for me to pontificate on that here 🙂

Pls let me know what you think. It’s such an exciting time to be a part of the tech ecosystem – I am sure there are counterpoints to what I said above and I look forward to learning from you if you share your thoughts in the comments

GenAI maybe better validators for now than creators


If the hype is to be believed, then GenAI is the answer to all questions these days, isn’t it? We liberally use GenAI every day even though we know that hallucinations are largely unavoidable.

Two of the most popular use cases in my line of work are

1. Code generation

2. Text summarization

None of us who work with GenAI tend to position it as “autopilot”. We know that it needs humans to be in charge. That’s why GenAI solutions are usually called “copilots” and “assistants” – it has fundamental issues on accuracy and reliability which need humans to rectify and sometimes just ignore. It could generate code that might be syntactically correct but not doing what it is intended for. It could write up a creative summary that makes up invalid data – like nonexistent citations and so on.

To be fair – humans make up stuff all the time too. We often act without thinking through stuff in any detail – but if asked to explain our action, we can usually come up with a convincing explanation. For example – I just returned from my walk. I could have taken a dozen different routes today and I have no clue why I chose the route I walked. But if you asked me – I could come up with some decent answers like “this is the route that has the least amount of traffic this time of the day”. This answer is a believable one for most people and they have no reason to believe that I just made it up looking backwards.

It is relatively straightforward to verify whether my explanation was factual or not if someone wanted to spend the time doing it – and it doesn’t need as much creativity as I needed to come up with a creative answer.

If we extend this concept back to GenAI – it’s not a stretch to see how it’s a lot more efficient (and quite valuable) to use the tools to validate code and text summaries, than creating them in the first place. It takes hardly any creativity to check if a citation is valid compared to creating a fake citation that comes across as realistic. Similarly – it’s a lot easier to create a comprehensive set of test cases for a given code base than creating the best code to solve a given problem.

When I explained this thought to a friend last week in India – the pushback I got was that he didn’t think a system that does a less than stellar job of creating code can be that good at testing.

I think this lack of trust is a bit misplaced

1. let’s say we are building a plane with the best engineers on the planet – half of them doing the build and the other half doing testing. Would we be satisfied with only the engineers who are qualified to build as testers? Or would we ask for engineers who are test specialists? And in any case will we trust it till a pilot actually flies the plane? The ability to build a great plane doesn’t translate directly to the ability to test it thoroughly – or vice versa. And there is no necessary constraint that the same person needs to be an expert in both

2. AI is a lot of more useful when boundary conditions are known – which is the case when all you need to do is validate specific things. In fact you can use a lot of deterministic techniques – and generally improve computing efficiency – when you have specific boundaries to the problem.

I absolutely think that over time GenAI will largely overcome its deficiencies on accuracy etc. But we don’t have to wait for that to make it useful for us – and validation use cases might be one such high value pattern.

I am curious to hear your thoughts about this. Pls leave a comment if you could.