There is no shortage of froth around AI (understatement of the year?), and to accompany it there is a growing frustration with all things AI: the pace of measurable outcomes, pressure to token-max, the avalanche of news, tools and workflows, and the urgent sense that "we are missing the boat."

At the same time, the use of AI in everyday work continues to grow. How many people do you know in 'knowledge work' who do not use AI in some fashion on a regular basis? For many leaders, that's great — we want people using AI in their work. The growing caveat is that we want them to use it in ways that deliver meaningful and measurable business outcomes for the company.

And herein lies the rub.

At the same time that more people are using AI to help them do their everyday work, the pressure is on for leaders to show that AI is improving business performance. I agree with the argument that the days of token-maxing are about to sunset (we can have a good and spirited convo on the yin and yang of that!), and the emphasis will shift to a focus on real outcomes.

Which will mean most organizations will need to take a step back and start at the beginning. Which, in turn, will mean not looking at the agents, tools, or workflows — but rather the outcomes: what does success look like? What business objectives are we supporting? What processes can we streamline, augment, or automate?

The question NOT to ask: what are the hot new tools or agents? Why? The technology is the easy part. The harder part is the thinking part. The strategy. The process. And most important — the people.

Most AI initiatives stall not from a lack of technology but from a lack of structure. Here's why sequencing is the variable that changes everything — and how to get it right.

Sequencing is not a framework. It's a discipline.

I am cognizant that "sequencing" can sound like consultant-speak for "do things in the right order," which is not particularly useful advice.

What I really mean is this: most organizations sprint to deployment before they've done the thinking that makes deployment worth anything. They sign the contract, stand up the platform, announce the initiative and then realize that the conditions for success were never in place.

The sequence that works is not complicated. But it requires slowing down at the start, which is the hardest thing to do when there's organizational pressure to show AI progress. Strategy first. Process second. People third. Technology fourth.

Every time you invert that order, you're borrowing against future rework. And in our experience, the debt comes due faster than anyone expects.

What has to be true before the tools go in

Before any AI initiative goes live, there are (at least) four questions to consider:

Do we know exactly what problem we're solving? Not a generic "improve rep productivity" – that's a direction, not a problem. The problem should be specific, bounded, and measurable. "Our AEs spend 40% of their time on post-call documentation and it's killing pipeline velocity" is a tangible problem AI can solve.

Is the underlying process clean enough to automate? AI doesn't fix broken processes. It accelerates them — which means a broken process becomes a faster broken process. If your lead qualification criteria are inconsistent across the team, AI will make that inconsistency happen at scale. Fix the process first. It's less exciting than deploying the tool, but it's the work that actually matters.

Is our data in a place where the tool can trust it? This is the question most teams don't ask honestly. CRM data, in particular, is almost always messier than leaders believe. Before any AI tool can score, prioritize, personalize, or predict — it needs data that's complete, consistent, and reasonably current. If it doesn't have that, it's guessing. And you're paying for it to guess.

Do the people who will use this understand why it exists? Not "we sent a training invite" but rather, do they understand the problem it solves, how it fits their specific workflow, and what's expected of them? If the answer is no, you don't have an adoption plan. You have a technology deployment. Those are different things, and they produce different results.

The conversation most leaders try to avoid

Here's what I observe in nearly every AI roadmap conversation: the team is excited about the technology, the vendor demo was compelling, and there's real energy in the room. Nobody wants to be the person who slows things down.

So, the hard questions don't get asked and the process gaps don't get surfaced. The data quality conversation gets deferred, and six months later, the initiative stalls and everyone is pointing at adoption.

The conversation that has to happen is a realistic assessment of where your GTM operation stands today, and what measurable business outcomes you are organizing to solve for.

That conversation can be uncomfortable because it surfaces gaps and misalignment, creates accountability, and sometimes reveals that you're not as ready as you thought. But it's also the conversation that separates the organizations that get real performance from the ones that get a lot of AI activity and not much to show for it.

Start with the question, not the tool

Most successful AI initiatives don't start with a vendor evaluation. They start with a question: what does performance look like in 90 days, and what has to change in our strategy, our process, our data, our people for that to be achievable?

The tools come last. Not because they are unimportant (in fact, they are critically important) but because the right tools, deployed into the right conditions, with the right people prepared to use them, perform. The same tools deployed without those conditions just creates a more expensive version of the problem you already had.

Sequence it right, and the technology does what it's supposed to do. Get the sequence wrong, and you'll spend a lot of time in a postmortem examining why your AI initiative flailed.