Introduction

I am a kind of, I guess, a serial founder. Well, I'm a second time founder.

I founded and sold the company a couple of years ago. We sold big banking, annoying software to huge investment banks. um so my background was basically uh building good relationships with these very complex organizations and uh selling and implementing very complicated software um i sold that in what in retrospect looks like staggeringly good timing at the top of the market um and instantly developed post exit midlife crisis and uh you know did all sorts of stuff land fly etc

Now, having got a bit bored and got very excited with artificial intelligence, I'm going at it again.

The Journey from Idea to Commercialization

And so what I'm here to talk about is basically how do you actually take an idea and commercialize it in the sort of B2B space? And I wanted to share with you some of the sort of benefits of our learnings, I guess.

Early Market Experience

We launched our product in December, so we've had four and a bit months in market, kind of quite seriously selling. And I wanted to just talk a little bit about some of the pitfalls and maybe spark a discussion about how best to avoid some of those.

The Misconception of AI Product Sales Simplicity

So for those of you who studied maths or computer science or something with logic in it, it kind of feels like B2B sales for an AI-powered product must be easier. It's almost like a sort of logical syllogism, right?

So if... it's possible to sell a normal B2B product. And if your product includes AI, then it's probably better. And if the better product is easier to sell, then surely selling an AI-powered product to businesses should be easier.

That was certainly, I guess, the naive way that we approached this. One of the things I was extremely excited about was how easy everything was going to be.

And, you know, it was very tempting in the early days to just pivot instantly into, you know, forget about building the rest of the product. Let's just build a, you know, AI powered counting bot to like count all the money that we're going to make. Unfortunately, it's a little bit more difficult than that.

1So in some ways, it turns out selling an AI-powered product B2B is actually a little bit harder than selling, you know, kind of in the old world. I think there's a couple of different reasons for that. And, you know, I'm only one proof point or I'm only one sort of sample point on your data set. So bear in mind, your mileage may vary as always.

And I think that also I may not have the most amazing solutions for you today that are distinct from solutions that you would use to solve any kind of sales challenge or overcome any sort of sales challenge. So some of this will sound a little bit like motherhood and apple pie, kind of good, sensible wisdom, but nothing revolutionary. So hopefully we'll dwell on some of the things that are actually more specific to the artificial intelligence arena.

Understanding the B2B Sales Process

So first, it's sort of helpful to lay out what the sales process looks like. So normally you're talking through a couple of different stages in a business sale. You're saying, people need to be aware of what you sell, otherwise they never come in the door. They never put their hand up and say, I'm interested.

You need to be able to understand what their problems are. You need to pitch your product to them. They need to somehow like evaluate that pitch or kind of get comfortable with it, de-risk it. And then you need to implement your product in their business.

And so unfortunately, it turns out there's sort of AI specific hurdles at each of these stages or AI specific complications. So I'll talk a bit more in depth about all of these.

Challenges in the AI Product Market

But at the awareness stage, there's just a lot of noise out there. There's a lot of people gabbing. There's a lot of people trying to sell.

Every buyer in a business is being pitched 900 different startups, and probably 80% of them these days at least have some sort of AI story to them. There's a lot of tourists as well. A lot of people in enterprises have been tasked with working out how they use AI in some way, shape or form. And so there's just a lot of people out there who are basically kicking the tires and stuff, but have no purchase intent and no actual interest in really, you know, paying your money.

Demo Challenges with AI Products

1AI actually presents some real challenges to what I would call demo-ability. While demos of artificial intelligence products can be really magical, at the same time they can actually be extremely difficult. to sort of convey the shit, the power of your solution.

It can be really, really hard to And I think that's definitely true for B2B apps. I think no one really has a buying framework for AI yet, and that makes it a lot more difficult.

I think in consumer, it's a bit easier. You can show people something which is kind of just, you know, magical for their use case. But in a B2B, you've got to show that it's going to actually fit into their business problem.

Implementation and Security Concerns

And inevitably, when you actually get towards implementation, security, privacy, all that stuff starts to be a real concern. So I'm just going to run through these quickly, and then we can have a couple minutes for questions if people are interested, the three or four people who actually want to commercialize their work. And then we can move on with the drinks.

Cutting Through the Noise

So I think the first thing is like, how do you cut through the noise? And I think that there's tons of challenges here. At what level should buyers be thinking?

Are they thinking about just like buying chat GPT for everyone, which is something I've heard from a surprisingly large number of big companies actually. you know, are they thinking about bringing in a new vendor? Are they just trusting Microsoft will like solve all this stuff in some like quasi-magical way? You know, they were already by Office 365 and like eventually everything will get integrated in a way that means they never need to interface with anyone else.

Or do they need to take it to that, you know, 25th meeting of the GNI steering committee, you know, which happens on, you know, the third Thursday in the month or something. at which the IT department pontificates about which solution they're going to divert their limited resources to next.

And I think that really here, it's all about hyper-specificity, especially at the early stages. Who is your buyer? Exactly what problems are they wrestling with? And this is sort of generic advice, but I think it's particularly applicable to these magical solutions with artificial intelligence is that you want to really focus on like what the business problem is as opposed to what your solution is and how it delivers it.

And I think we fell foul definitely initially of like talking about ourselves as primarily AI enabled rather than just really focusing on how we were going to enable businesses to implement a really, really efficient and effective help desk really quickly, right? And the fact that it's AI enabled means you can actually do it as opposed to it being rubbish product. You know, we sort of led with the technology in a way that I think in retrospect was naive.

Qualifying Prospects and Learning from Customers

At that second stage of the process, how do you sort of really qualify out your customers? And here's maybe the place that, not qualify the customers, but how do you qualify out bad leads, bad prospects? And I think it really is just about relentless qualification. You've got to kiss a lot of frogs, I'm afraid.

You have to at the early stages, I mean, this is again, mother of an apple pie advice. You really have to optimize for how much you're learning from your customers rather than on shepherding each of them through to a successful account, unless you want to build a consulting business, which is fine. Um, but you know, in order to get a product that actually stands on its own two feet and has a target audience, you need to be really focusing on how do we learn as much as possible about the consistent thread that runs through the prospects and what they're actually willing to pay for.

The Art of Demonstrating AI Products

pitch is one of the areas that i think there's a lot of ai specificity there's a lot of big challenges in in demoing and in communicating the value of artificial intelligence driven products one of the things we've learned is that a product that we think is magical to use because we have all the context on the data that surrounds it on the way that it processes that data on the confidence that we have around the results because we've used it hundreds of times and we've really experimented with all the different permutations of how it could go wrong and we fixed all those things.

When you have an opportunity to show that to a customer in a meeting of 15 minutes, 30 minutes or something, it's asking them to take in a vast amount of data to actually make sense of what you're proposing to them, what you're pitching them. I guess if I could sum this up in a couple of words like Chatbots demo very badly. Things like dashboards, like things with colorful graphics, demo extremely well.

You may even find yourself having to build features that you don't expect anyone to actually ever use in order to communicate the value of the product that you do have in a way that's succinct and easily memorable. Because otherwise, all someone really remembers is just a wall of text.

It's really difficult to sort of empathize enough with the customer that you just have to keep coming back to basics and keep trying to remember that when you show them, I don't know, 60 words, a sentence that's been generated, whether it's in the context of a report that you write or a message that you would send or a draft email or whatever it is, The customer can't process that like no very few people are able to sort of process that and even fewer able to really put that information in the context of Their business problem in any meaningful way so I think this is an area that if I would hope that some of you will go away from this and think actually maybe you know we can really think about how to improve the quality of the demos that we we've created for for these products and

Evaluation Stage: De-risking Customer Decisions

I think that at the evaluation stage, we're dealing with an immature, well, we're at an early stage of a platform shift, right? And customers don't really know how to buy yet. They don't really know how to, the buying process for them, for a B2B customer is about de-risking their decision from their point of view in terms of like, will this product do what it says it will? And am I getting, is it gonna go wrong in a way that embarrasses me or causes me trouble?

And people just don't have the assessment frameworks. They don't have the ways of kind of evaluating products in a consistent way that allow you to control the sales process as a vendor. And so what happens is that people will go for comfort.

They'll go for things like trials and trialing a product in most B2B contexts with a centralized buyer as opposed to product led growth ones where you're sort of selling into a team and then growing the account up from there. tend to be very unsuccessful because basically you're asking people to sort of like kick the tires on something without really committing to it and in general most products don't succeed very well in that type of environment normally successful b2b sales processes work by you make a customer fess up to the problems that they have in their business they decide to adopt the solution you're proposing and then you jointly work on how to make that solution effective and And so putting them in a position where they're trying to do some sort of ad hoc assessment of the quality of your product is actually just like asking them to fail, basically. It's asking them to undertake a task they're not equipped for.

It's extremely difficult as well in an AI context to do things like validate that your product doesn't hallucinate. You just basically end up with an army of slightly hostile QA testers who never give you proper feedback and never actually allow you to improve your product. They're just making your life difficult and frankly wasting your time. So controlling that is really difficult.

The best way you can do it is by sitting down at the outset and agreeing an evaluation framework, agreeing what the success criteria will be and making sure that you're able to take the sting out of any kind of negative risks. For example, by offering things like guarantees rather than trials, you can reframe it into a kind of default yes outcome rather than a default no outcome.

Compliance and Security Measures

And I think the final thing, after which I'll shut up and we can all have a beer, is again, you know, quite a sort of generic piece of advice, but I think it's increasingly important in general and in particular with the AI wave, which is that unfortunately you just have to do the boring work around compliance, especially things like SOC 2 ISO 27001, I regard it honestly as security theater on one level or another. I think it's good stuff to do generally, but it doesn't make you a secure business to have ticked your boxes around access controls, et cetera.

But I think that every day we hear customers asking for commitments around where is your data going, what data is used to train models, is our private customer data used, is our employees' private data sent to models, is it stored, et cetera, et cetera. And we're seeing this even from businesses with 50, 100 employees that don't have a proper procurement department, that don't have a level of sophistication, frankly, to evaluate smaller vendors. Ironically, it's the bigger purchasers that are in a better place to actually ask you meaningful and sensible questions and successfully shepherd you through a procurement process.

We've certainly found that for us, you know, we're a business of four people today, having SOC 2 and having ISO 27001 is actually a must even at this scale, which like, frankly, in my last business, it was never a big deal. And we sold to like Goldman Sachs and Citi and JP Morgan. Like we never even bother doing it. So that's partly just a sign of the times, but it's also, I think, the level of sensitivity around the data that goes into models and also just the generic anxiety around what the models are actually doing.

Conclusion

So anyway, I wanted to say, I wanted to sort of raise those things and say thank you for letting me talk and offer myself up for questions.