Put AI behind your sales reps, not in front of them

Every sales team will eventually have AI agents running behind it. We are fairly sure of that. Gartner expects AI agents to outnumber sellers ten to one by 2028. What we are less sure about is how many teams will get there in a way that adds value. Gartner puts a number on that too, in the same release: fewer than 40% of sellers will say the agents improved their productivity.

That second number is the interesting one. It says the agents are coming either way. It also says most of them will land badly, built onto a messy process, adding more tools and more prompts to a week that already had too many. The teams that get value from AI agents will be the ones that arrive at them in steps, with each step paying for itself before the next one starts.

This is how we think about those steps and why the order matters.

Efficiency is the small win

Most conversations about AI in sales are about doing the same things faster. Write the follow-up in two minutes instead of twenty. Get the call summary without typing it. That is real value and it is where we start with every team.

The bigger shift is that AI agents can come up with processes nobody on the team designed. Give an agent a goal (keep every open deal moving this week), some guardrails (never contact a buyer directly, never change a close date without a rep signing off) and access to the CRM, the inbox and the call notes, and it will find its own way to get there. Some of what it finds will be things your best rep already does without telling anyone. Some of it will be new.

You cannot skip to that. An agent working from a process that only exists in people's heads has nothing to stand on.

Step one: write down how you actually sell

Ask a sales leader how a deal moves from first call to signed contract and you get the version on the slide. Ask three reps and you get three different answers, all of them true. Most of the process lives in the heads of the people running it, plus a few Slack threads and a CRM that everyone fills in differently.

So we begin with an audit. We sit with the team and map how sales runs today: what happens after a first call, who writes what down, when the follow-up goes out, what a rep does before a demo, what "qualified" means in practice. Not the ideal version. The real one, including the parts that work and are worth protecting.

That map is never finished after one session. It is the first version of a blueprint and we keep building on it. But even the first version does something useful on its own: it shows where selling hours go to the work around the deals rather than the deals themselves. Salesforce finds the average seller spends about 40% of the week selling, and the rest goes to everything around it. In our experience, once a team can see that on paper, nobody argues with it.

Step two: automate what is already there

With the map in hand, the first automations are obvious. Call prep before every meeting, pulled from the CRM and the last few emails. Follow-up drafts written while the call is still fresh, in the rep's voice, waiting for review. CRM updates that happen because the call happened, not because someone remembered on Friday. A pipeline review that lands on Monday morning with the stale deals already flagged.

Do not build all of it at once. Make a shortlist, starting with what gives the most time back for the least effort to set up, and work through it one by one.

None of this changes the process. Each automation runs one part of the existing process in the background, without the rep doing the typing. That is the point. Reps get hours back straight away, on work they never wanted to do in the first place, which is why this step rarely meets resistance.

It also does something quieter. The team learns what AI is good at, and what it is not, on their own deals rather than in a demo. They see where the draft was right and where it missed. They start trusting the output where trust is earned and correcting it where it is not. That learning is what makes the next step possible. A team that has never had AI run its admin for a few months is not ready to hand an agent a goal and let it go.

Step three: agents that find their own way

Once the process is written down and the admin runs on its own, you can start building towards agents that come up with their own processes. This is where it gets interesting, and where the guardrails from step one earn their keep. But at the same time, this is a longer term step.

An agent given the goal of keeping deals moving might notice that deals stall after the second meeting and start preparing the rep for the third one differently. It might spot that a buyer has gone quiet and draft a nudge based on what worked in similar deals. It might find that the CRM stages do not match how deals really move and suggest new ones. The rep did not design any of those steps. The agent did, inside the limits the team set, and the rep still decides what goes out.

Looking back at the three steps, the one most teams skip is the first. Somebody sees a demo, gets excited, and a week later there is a tool writing follow-ups. But without the blueprint, the choice of what to automate is a guess. Teams pick the task that was easiest to see, or the one the vendor showed them, rather than the one that was costing them deals. Then they build agents on top of that. Agents scale whatever they are pointed at, so a guess at step two becomes a guess at scale in step three. That is how you end up with more digital activity and sellers who feel no better off, which is exactly what Gartner is predicting for most of them.

Behind, not in front

The line we keep coming back to is that AI belongs behind the rep, not in front of them. The rep still owns the deal. They own the relationship, the conversation, the judgement about when to push and when to wait. That is what they are good at, and it is what buyers want from them.

AI does the rest: the prep, the follow-up, the CRM work, the pipeline hygiene. The admin. In time it also does the thinking about how that admin should work, but it does that thinking behind the rep, and the rep stays the one who sends.

If you want to know where your team sits in that sequence, the first step is the one we can help with straight away. Our free audit maps how your team sells today and shows where the hours go. It is the same map we would build in step one, and it is yours whatever you do next.