Why most AI projects fail (and how yours might not)
A conversation with Saurajit Kanungo, CG Infinity President
Nathaniel Rounds, OfferFit Product Marketer: So you are President of CG Infinity, which is a technology consultant and implementation partner. Could you say a bit about what your business does?
Saurajit Kanungo, CG Infinity President: Yeah, happy to. Businesses across every industry are constantly evaluating and buying different technology platforms. They could be investing in a customer portal, or a mobile app, or a CRM system, or maybe a campaign management system. A technology consultant like me, or an implementation partner like CG Infinity, helps our customer implement those systems into their existing ecosystem of technology. That means managing the project, integrating, making sure that their users adopt, and making sure they gain business value. Our industry is rather young, but growing fast.
Nathaniel Rounds: Why do you think the tech consultancy industry is growing so fast?
Saurajit Kanungo: I think the number one driver is the increasing realization that every business needs to be a technology business. The second factor is the rate of change in technology – it’s so fast that for an internal IT department to keep up, that’s just going to be impossible.
Nathaniel Rounds: And how should companies be thinking about getting value from all this technology, all the data that they have?
Saurajit Kanungo: Yeah, this is a great topic. There are zettabytes of data being created every year. IDC reported that the amount of data being generated will grow by 21% a year. Clearly our digital storage is not growing at that rate, so we have a physical boundary. In spite of that, we see that only a very small percentage of best-in-class enterprises are making significant investment in data. They all know that data is the next oil, right? But only a very small percentage of enterprises are able to extract the data, refine the data, and leverage data for driving business value.
I read a report in Inc magazine not long ago that 73% of the enterprise data gets unused. We see a very small percentage of enterprises that are trying to make the jump to unlock the value of their data. But we also see that in a year or maybe 15 months, they're having to pull the plug on these data projects. And it's really unfortunate because I think we all will agree that there is tremendous opportunity. There are examples of where a small set of enterprises are actually unlocking value. But how do you make sure that the percentage continues to increase? That's a huge opportunity for everyone.
Nathaniel Rounds: So when you talk about a year, 15 months, this project gets cut, what do you think is driving that? Why are people starting to make the right steps and then backing off?
Saurajit Kanungo: Well, I think a lot of different reasons. Enterprises tend to put the tool and technology in the forefront. “I want a neural network.” I think the business value takes a backseat. I also see people want to go after a grand slam on their first at-bat. It's a situation of the perfect being the enemy of good, I would say.
Fortune magazine reported that the failure rate on AI projects has been between 83% and 92%. And if you go to Vegas, I think you’re keeping more than 10%, right? So if you're a business person, you say, “Okay, I burned my fingers, and there's an 80% to 90% failure rate. Should I roll the dice on AI or should I go to Vegas?”
Nathaniel Rounds: If I were one of your clients and I said, “I just read this report that 73% of companies don't use any of their data. I feel a lot of guilt and shame about this. I don't think we're using our data, but I'm afraid of getting burned.” What would your advice be?
Saurajit Kanungo: I would say three things.
Start with some low hanging fruit. Let's look at your business environment. Do you have competitive pressure or do you want to go after new markets?
Second, start small and get quick wins. Let’s not try and build a bridge to nowhere.
Third: measure, learn and evolve.
So let's start with a very low hanging business goal that we can measure, and let's keep it small. Then let's learn and do it again. That would be my advice.
Nathaniel Rounds: Can you give an example of what this might look like? What might you recommend as the first bite of the apple in terms of data science investment or machine learning investment?
Saurajit Kanungo: I will actually give two examples.
One is our partnership with OfferFit – what we offered to one of our mutual customers. I think it's a great story. It may seem like an AI project, but to our mutual customer and to me, the AI was the means. AI is not the end. What your project is doing, and did, is drive customer retention. So it's increasing revenue. And we didn’t ask our customer to wait 15 months to see a result. We just said “let's drive customer retention and measure our progress as we go.” So this is a great example of starting small by hitting a single. I think we gave our mutual customer a few grand slams, but we started small and got a quick win.
The other example is a customer in real estate, a few-billion dollar company down in Houston. The CEO is a good friend of mine and she says, “We collect a lot of data, and we should invest in AI.” And my first question to her was, "Why should you invest in AI?” She was puzzled because she attends the workshops and seminars that we conduct about unlocking the value of data – shouldn’t everyone invest in AI? So I said, “I think we need to find the low hanging fruit.”
We realized that they have tens of thousands of homes that they own and rent out. We found out that the rental prices are based on history and experience. So I said “hey, how about we do a quick POC on this? It'll probably take no more than four months, but let's put some very simple statistical algorithms to show you recommendations for pricing.” And you won't believe it – in four months the result was that they're probably leaving somewhere between $800K to $900K worth of incremental rent on the table.
So let's start with the business value. Let's not start with, “Well, you know, we need machine learning or neural networks.” Don't start with the hypothesis that this is powerful technology, so it’s got to be able to do something for me. Business value and quick wins need to be the starting point.
Nathaniel Rounds: What if somebody comes to you more in the middle of this process of adopting AI, and says, “Saurajit, I have all these investments. I use a CDP. I have an orchestration platform. I don't know what I'm getting from it all.” How would you advise in that kind of situation?
Saurajit Kanungo: It's a really hard problem to solve.
We're fortunate in that at CG Infinity we probably execute over 100 projects every year. And whether the customer asks for this or not, we want to somehow measure value. We’ve become very disciplined about it. We want to measure value on every project we are delivering to our customers. And then the question is, what KPIs? So we boil it down to three simple KPIs:
Reducing cost, and
Every project we do has one of those effects, and we present that in our monthly relationship meeting and quarterly review meeting. It keeps my team disciplined. We’ll talk our customer out of a project if we are not able to say what the value is. In terms of where I want the maturity of our reporting process to be, we are probably five out of ten. We're not there yet, but we're absolutely driving that process internally.
Nathaniel Rounds: To what extent are customers evaluating technological tools themselves along these lines, to quantify their ROI?
Saurajit Kanungo: Not much, really. And they actually show some hesitancy when we do it for them. But we all know everyone has a boss. What I have started to see is a few of our customers, let's say the CIO, starting to use our quarterly report with the CEO or CFO. They're able to defend their spend, or defend why they are doing a project. So even though customers don't specifically tell us that's why they need our analysis, probably 20% to 30% of our C-level customers are using our reports to justify the projects that they are undertaking.
So we’re sort of modeling good behavior for our CIO customers. We’re giving them a model of how they can explain the ROI of their investment. And it's not purely Good Samaritan – we know that as vendors, when the gun is pointed at the CIO, more than likely the CIO will deflect it to somebody like us. And so measuring clear ROI is also risk avoidance and maximizing our CLV with that customer.
So I think the trend towards clearing measuring business value of tech projects is going to continue. As technology professionals, we need to make ourselves accountable. It may not be a very scientific process; it's probably a more heuristic process.
But I think it's the right thing to do.
Ready to make the leap from A/B to AI?