Let’s start from the top here: when should founders consider going AI native for their revenue teams?

To get AI working, the leadership team needs to be realistic about where you currently are. Throwing AI at broken GTM motions, broken systems, roles that aren’t working out, people that aren’t working out is not going to fix those problems. You have to have something that’s fundamentally working in order to use AI for most cases.

However, it is okay to use AI in situations where you don’t know quite what the answer is and you want it to help you synthesise large amounts of information in order for you to make a final decision. So there are some circumstances where things aren’t already working that you can use AI as a thought partner, in order to help you work out the best path forward.

For example, you might have multiple options in terms of where you should be targeting your product and who your ideal customer profile is. That is ultimately a decision not to be made by AI. It has to be made by, ideally, the CEO or the founder. The buck stops with them. But where there is an awful lot of ambiguity or lots of data, perhaps contradicting data or stories from the market around where the best fit is for your product, then you can gather lots of insights from doing massive amounts of external research, from looking at all the data in your CRM, from digging into all of the calls that you or your reps have had. These are the types of mass volume work that you simply don’t have the time in the day to do. And having AI run through it all, synthesise it, create context, categorise it, so you can take that and make a human judgment call on top of it, is a pretty good use case.

Otherwise, what I would say is that you want the foundations in place, such as knowing roughly what your go-to-market motion is. So you know whether you’re PLG, SMB, mid market or enterprise, you have a pretty good idea about what your ICP is, what your value proposition is, who the personas are that you’re going to sell to, and why they would buy from you versus others. These are some of the foundational elements. Again, I’d encourage anyone to think about Brian Balfour’s ‘GTM four fits’ and whether you believe you have those.

Broken GTM versus genuine ambiguity

The difference between broken GTM that AI can’t fix and ambiguity that AI can help with comes down to whether the founder or the person who own the problem genuinely knows what the parameters are. Do you know what you’re even trying to ask, or the answer you’re trying to get to?

So, for example, if you find yourself thinking “We need more pipeline, and we are sending thousands of emails a day and not getting a response, so let’s fire the SDR team and bring AI in to do the same thing, but cheaper, faster, at greater scale.” It’s not going to fix things, because you haven’t sorted out the fundamentals, like who are you targeting, when are you targeting them, with what kind of message and value proposition, what’s the offer, what’s the CTA, and so on. And AI isn’t going to do a good job of fixing all that.

However, if the situation is more like this: “Okay, we’re not sure why this isn’t working. Let’s use AI to try and figure this out.” And what you did is use AI to look through all of their past communications, scrape all their calls and synthesise them, and look at patterns around successful accounts that have responded, and opportunities that have opened versus those that haven’t. You’d try and find the patterns that would suggest how you might update those fundamentals, such as who should we be targeting, when should we be targeting them, with what messages and so on. Then AI could potentially help you resolve those core questions in a competent way.

How to organise and operationalise launching AI into your team

Top down, after bottoms up discovery

In my experience, successfully implementing AI that gets to a state of being production ready, and that isn’t going to distract the team but actually add real business value, means you should be taking a top down mandate when it comes to the build and execution of it.

That doesn’t mean you don’t first speak to the reps and the other individual contributors that AI is going to touch, and understand what they care about, where they’re struggling, what their challenges are, and what opportunities they see. So you do need to do some initial groundwork to diagnose and understand where AI might have the biggest impact. But you shouldn’t go and give them all a budget and a mission to build stuff in AI and use AI as much as possible. That is a surefire way of burning through tokens, distracting your team, and probably having a ton of overlapping or relatively useless AI within the organisation that ultimately gathers dust and doesn’t drive real enterprise value.

So I do think you need to go top down after doing a bottoms up discovery and diagnosis exercise.

Agree the metrics for success

The next thing is to agree on the metrics for success. Implementing AI for the sake of AI without knowing what you’re trying to achieve with it is also typically a recipe for disappointment, and for the CFO to start putting the brakes on it at some point in the near future.

So what you want to do is look at a business level. What are the weaker metrics? That means the ones that are at risk of not hitting their number. It could be that you don’t have enough pipeline coverage, or not enough pipeline is converting from one stage to the next, or close rates are down, or deal values are down, or sales cycles are increasing significantly. All of those kinds of sales velocity metrics that mean it’s harder to hit the growth number, those are where you should be looking.

And then, normally, you need to break those down further and look at the inputs and KPIs that are driving those sales velocity metrics. That’s where you then want to tie AI initiatives to them, saying, “If we do this, we expect this number to improve.” That way, you can start to see a measurable result.

Give it a designated owner

It doesn’t really matter who owns it per se, so long as there is a designated owner. Obviously, if you have a GTM engineer, then yes, that makes sense. Otherwise, possibly RevOps: someone who’s quite operational, systems thinking and hands on is a good person to own the implementation. If you’re working with an external agency like AI GTM Studio, then we’re the ones that are responsible.

And then there still needs to be a good executive sponsor, who’s typically going to be the VP of sales, VP of marketing, or the overall CMO or CRO.

Use cases that tend to be successful

Once the foundations are there, AI can be used to either accelerate the things that are working and make them more efficient and scalable, or it can help you introduce new capabilities that you wouldn’t otherwise be able to deliver with the current resources you have at hand. For example, potentially running a more scaled ABM campaign, launching one, or moving into enterprise marketing.

The third would be helping you make decisions by taking massive amounts of unstructured data and helping you parse that into more manageable decisions, based on the context it provides you. That could be everything from switching up your ICP, to which markets to tackle next, to what products or features we should be launching. Maybe it’s down to which accounts within our overall TAM should be our focus accounts for this quarter. The final decision does, again, have to rest on humans, but the context, and the ability to present that in a meaningful way to the decision makers, is a great use case for AI.

And then having it run through your systems, so that it’s capturing and categorising data and execution layer data as it’s happening across the organisation, storing it correctly, and then serving it back up to the right people at the right time in the right places, is another wonderful use case in order to help your team do better work with better context, in the places where they’re already productive.

Examples from my client work

Some examples from my work:

• Freeing up rep time to make more calls. I worked with one business to help them become more productive in building pipeline by freeing up the sales reps’ time to get on the phone more and make more calls. So we built a series of AI skills to help them find contact details and context faster, so they weren’t context switching and moving between platforms, which had massively cut down their dial rate. With them, I also helped build a cold calling coaching skill to further develop the muscle and help book more meetings from cold calls. So that was a pipeline metric.
• Booking more meetings within a limited TAM. Another pipeline project was helping a team book more meetings with a limited TAM. They had to be very focused, so we built a custom signals platform in order to identify the right accounts to reach out to at the right time, with the right context and quite a complex, unique hypothesis around what might be driving those business changes. That led to more meetings being booked with those target accounts, and therefore increased pipeline and the quality of that pipeline.
• Overhauling inbound workflows. Another was around completely overhauling the inbound workflows of a team. In this case, they were able to significantly reduce headcount on the inbound side of things and essentially allow their existing AE team to deal with all the inbound volume, because it was being properly triaged. A few metrics moved there: an increased number of meetings held for AEs, decreased time to first touch for those inbound queries, and ultimately, again, more pipeline being built.

Those are just some examples.

The most common mistake

One thing is that teams probably overestimate the capability of AI and underestimate just how much time and energy and effort and input and context it takes to actually wrangle AI from the initial idea through to something that is actually production ready.

So often, it takes seconds, minutes or hours to build what looks like a pretty capable and exciting prototype. Everyone gets really impressed, and they think, “Oh my god, I can’t believe I was able to do this so quickly.” But the moment you try and get that to production grade, so that the whole team can use it and it’s not having bugs and it’s not hallucinating and it’s running as expected on a consistent basis, you realise that actually the last mile, the final 5%, even the final 1%, is often two, three, 10 times harder than it was to get the initial prototype up.

And so what then happens is people go straight from the peak of excitement through to the trough of despair, and they don’t have the time or energy or skill set and experience to push through that. So you end up with lots of failed prototypes and projects that never actually deliver business value.

So it’s probably overestimating the ability of AI and underestimating the challenge of how difficult it is. But when you do make it production ready, AI moves the metrics that matter, the way it did with dial rates, meetings booked and time to first touch in the examples above.

Time to go ‘AI pilled’?

If any of this resonates, and you want AI to drive real business impact across your GTM team (and understand it isn’t some miracle pixie dust you can just sprinkle on broken processes), then I’d love to hear from you and we can kick off a free discovery session.