Role Of Services In Artifical Intelligence

Please don’t interpret the following as IBM’s position – this is just my view . Also, after a long time, I am typing a blog post on my mac instead of on my iPhone 🙂

I am not an official spokesperson for IBM . That said, I am not a spectator from the peanut gallery either. Almost every day for last couple of years, I have been involved in the sales and delivery of projects that include some flavor of Artificial intelligence. I am sticking with AI here instead of “Cognitive” to avoid the distraction of what is AI, ML , Cognitive etc . While I do not have an academic degree in AI, I have hands on design and coding experience in this domain, and I am learning more every opportunity I get.

Couple of weeks ago, I was in Vegas for a week – attending IBM’s World of Watson event. I spoke with tens of clients and prospects, several analysts and a bunch of my colleagues from other parts of the world that I don’t get to interact frequently. One of the common themes in this conversation was the role of services in AI. Variants of this topic also caught my attention on social media as well – mostly from analysts. The big difference between the people I spoke with was that many analysts were dismissive or pessimistic about the role of services, and clients did not seem to care what was product and what was services – they did not seem to care much about the distinction. Either way, this topic picked my interest bigly 🙂

There are many reasons to be pessimistic about the role of services in AI

  1. Things that caught the imagination in past like ERP, data warehousing etc all had big services components and many projects were plagued with cost over runs. Why repeat the mistake with AI ?
  2. Most AI initiatives – from IBM, Google etc – are cloud based. Cloud should not need as much services, right ?
  3. AI in general comes with a cloud of uncertainty about the future of mankind. Almost every person I know has a strong position on Universal Basic Income thanks to this fear. Why accelerate the apocalypse by promoting services on AI?
  4. AI is in its nascent stage. Services is what hides the immaturity of the product and hence deserves a public take down. In any case, it is more cool to be on the side of product and be opposed to services.

If you are running out of patience already on what my take is – I believe services will be integral to the present and future of AI, for a long time . Here is why ( and its long – get a beer or coffee ).

The nirvana state of AI for me is when computers can mimic ( and accelerate) human thoughts instead of just human tasks. We are a long way away from that stage. But what is possible today is already good enough to help individuals and businesses in tangible ways – often with “order of magnitude” benefits. People who criticize AI as immature often do not recognize that many companies use it heavily today, and its a part of many apps on your phone. The future is already here, and its getting more evenly distributed as well 🙂

A lot of people think – and I get asked almost every day – that systems like IBM Watson are big monoliths like traditional ERP systems. While the start was certainly that way , AI today is largely a set of APIs that application developers can mix and match for their unique needs. These are quite well defined APIs and even someone like me who is not a full time developer can figure out how to use it pretty quickly.

One such API is speech to text . You input the audio, and you get text as output – that is pretty much it. There is no end to the useful scenarios this “simple” API can be used in ( closed captioning in TV for example). However, there are very few people relatively who know such an API exists, and even fewer who know how to integrate into existing workflows in their business to improve efficiency and effectiveness of the process. Consultants can help their clients explore how AI can be incorporated into their existing processes. This kind of integration work is perhaps the most tactical use of AI services today.

While tactical uses are awesome, most companies have little interest in things that make incremental changes. I have a client who is in the entertainment business. The leader in their industry is about 4 times bigger than them. They are betting on AI as the way to leapfrog the leader and become the top dog in their industry. A good part of that project is the strategic planning – what bets to make, what trade offs to consider, how to mitigate risks etc. AI Advisory is probably the most exciting part of the future of consulting business. It needs a unique mix of AI knowledge , the intimate knowledge of specific industries, and business process expertise. In other words – product is not the lone differentiator.

I explain this to my clients as “It is not what we make with AI, it is what we make possible with AI. We will help you build the bridge from where you are to where you need to be” . 


There are some “tangential” scenarios too for AI in the Enterprise IT world. Between ERP and Data warehousing, countless days of my life was spent in getting data cleansed and transformed. For the most part, even with SaaS and big data – we worry a lot about GIGO ( garbage in garbage out). AI will help us take a fresh look at this problem. What we call as “garbage” data might contain many useful nuggets. We might not need to transform it row by row to get meaningful insights – AI can figure out useful patterns. Coupled with the modern big data management systems, this could make a lot of ETL disappear for scenarios where directionally correct insights are more important than precise information. It needs services – the kind that generates actual value and gets rid of a lot of terribly inefficient code. Actually not just code, also the precious human effort wasted in fixing the “garbage” data !

This post is already becoming quite long, so I will stop with making just one more topic – a favorite for both my clients and my analyst friends. Companies around the world have invested heavily in ERP systems ( and satellites for SRM, CRM etc). Those assets have not all been sweated sufficiently quite yet. Some of these vendors have modernized ( mostly about moving to cloud, and making the systems faster ). However, in the promise of “better, faster, cheaper” – the “better” part is largely unfulfilled today. I firmly believe the missing ingredient is purposeful use of AI.

Lets take the mundane case of managing collections – where optimization today means you outsource it and make it someone else’s problem. Cost of collection is not trivial for most clients I have met – but they have mostly just accepted it as a fair cost of doing business. Efficiency is improved by mostly fine tuning the process  part of it – like how many dunning cycles need to be run, but with very little personalization. We recently implemented some elementary AI ( trade off analytics, speech to text , text to speech etc) for a client who is a giant in the world of education ( took less than a month to implement it) and the client saw 20% better collections immediately after that.

Optimizing collections is really a trivial use case (and even that has tremendous ROI)  if we look at the power of AI available today. Exponentially better implementations will be possible (and already in the works) when we combine AI with other emerging stuff like say blockchain .

While I have no doubts that services will play a major role in the field of AI – I will be the first to admit that practitioners have a lot of learning and retooling to do to add value to their clients. The good consultants have always been life long learners – what is perhaps different for them is just the speed at which they need to learn now. Its a fast evolving field and there is just no time to sit back and say “I am an expert” any more.

PS: If working in AI, integrating cognitive capabilities to existing systems, Advising clients on making use of AI etc are things that you are good at, and you promise to keep learning – I am hiring. Ping me at Vijay dot Vijayasankar at US dot IBM dot COM . 

Published by Vijay Vijayasankar

Son/Husband/Dad/Dog Lover/Engineer. Follow me on twitter @vijayasankarv. These blogs are all my personal views - and not in way related to my employer or past employers

4 thoughts on “Role Of Services In Artifical Intelligence

  1. Any thoughts on how the planning phase of an AI project differs from the planning phase of an ERP project?

    In my opinion, not enough time is spent in the planning phase of an ERP project, many times because the service providers hypnotize prospects with some kind of “best practices” voodoo, giving them an excuse to make the sale quickly and then jump into billing.

    It seems to me like the number of options that you have with APIs is exponentially greater than the number of options that you have when boxed within the walls of a given ERP platform.

    It seems like one would have to spend more time in the planning portion of an AI project just to nail down the software price, given the potential for widely varying SaaS pricing, depending on which APIs are chosen and from which Vendors.

    If this is indeed the case, I believe that many service providers will need to learn how to turn their sales cycle into more of a consultancy cycle that they can get paid for. This would be a good thing in my opinion because now service providers would be getting paid for the portion of the project where they can add the most value.


    1. AI will eventually have its fair share of prepackaged apps like ERP – but probably with smaller purpose built foot print compared to ERP . ERP assumed what the vendor delivered was best practice . AI will learn over time . So I think at least for now that these two things will be fundamentally different


  2. Took long time to come, very useful post. I think, big part of services would be about educating customers about what is possible and how. I see that its on back of the mind , being talked about, but still largely in POC stage at most of the clients. Reason being, client have to do more heavy lfiting in AI projects than typical ERP implementations.
    Have you seen any large scale AI projects rather than just small scale ‘collections’ type ? Please share key learnings.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: