What a rollout involves
Rolling out AI sales coaching means getting a coach configured to the team's method, cleared by the people who review data, tested on the team's own cases, and then used weekly by managers and reps on live deals. One owner runs it, usually the head of enablement or a sales leader, and a first team can be live in about six weeks.
- Week 1Name the measure, appoint an owner
- Brief reviewers early
- Weeks 1 to 2Configure and test
- Configure to the team's method
- Test on your own cases
- Weeks 2 to 3Start with the managers
- Publish what managers can see
- Weeks 4 to 6Pilot with one team on live deals
- Week 6Review and expand team by team
Many rollouts are planned like a training launch: a kickoff session, a library of scenarios, a target for completions. That plan fades for the same reason training fades. In our field study of sales kickoffs, of about 4,800 first-hand accounts, 9 people described using something from the event afterward. What happens back at work is more strongly tied to whether a skill is used than what happens in the launch room (Saks and Belcourt 2006; Blume et al. 2010). So the plan below spends its effort on the weeks after go-live. For what an AI coach does on a live deal, see the AI sales coaching guide.
Choosing the coaching tool before the rollout
In our experience, rollouts that stall often picked the tool before agreeing on the problem. Settle step one of the plan below first, then test candidate AI coaching tools against it with these questions.
- Does it coach on live deals or only on practice? If the problem is stalled deals, a tool that only runs scripted practice will not reach it.
- Can it run your method? Load your framework and stage criteria, then check that the coach names gaps in your language.
- Is the deal review a call? Reps should be able to talk a deal through by voice, with typing as the fallback for when they cannot.
- What does the manager see? A short read of each deal and the gap it was coached on. Anything that needs a manager to replay recordings will not get used.
- Can you test it on your own calls? Grade a few past calls yourself, have the tool grade them, and compare before any rep sees it.
Treat the answers as the brief for step three, configuring the coach. A tool that cannot be configured to the team's method will make every later step harder.
The plan: eight steps
1. Name the problem and the measure (week one)
Write one sentence on what is going wrong in the pipeline and one number that will show whether coaching helps. A good problem is specific: "Deals reach proposal without a cost on the problem or a named economic buyer." A good measure tracks the deal: the share of agreed next moves that happen, or the share of late-stage deals with the evidence the method requires.
What good looks like: the sponsor, the managers and enablement agree the sentence and the number before any vendor demo. Logins and completions are activity counts; leave them out of the success measure.
2. Appoint one owner and brief reviewers early (week one)
One person owns the rollout from approval to the second team. Without a single owner, each reviewer waits for someone else to go first. Send security, legal and IT the vendor's documentation in week one, before they ask.
What good looks like: reviewers have what they need in the first week, and the timeline is set by the plan. For our documentation, see security and how we use your data; any vendor's claims should be checked against its own agreements.
3. Configure the coach to the team's method (weeks one and two)
Give the coach the method the team sells with, the materials that go with it and the products it needs to know. Where a methodology partner trained the team, involve them: they know what good looks like at each stage.
What good looks like: the coach asks the questions the team's method asks, in its terms, from the first session. A coach running a generic rubric gives advice reps have to translate, and they stop bothering (methodology coaching vs communication coaching). If the team uses Gap Selling, the ValueSelling Framework® or MEDDPICC, the coach should run it on the deal.
4. Test the coach on your own cases before reps see it (week two)
Pull a few dozen real coaching situations from the pipeline, have two or three methodology experts write what a strong answer contains, and judge the coach's answers blind, in pairs, against an alternative. Judge on how likely a rep is to act on the advice.
What good looks like: the experts would be comfortable putting the coach in front of their own reps. The method, and the failures it guards against, are in how to benchmark an AI sales coach. One finding worth knowing before this step: in our research on AI call scoring, a redesigned system prompt moved up to 40% of scores on a re-run. Run your test more than once.
5. Start with the managers (weeks two and three)
Before reps get access, each manager in the first team uses the coach's manager view (with us, Bryce's read of the team) for their own pipeline review for a week or two. They see which deals are missing which evidence and run the review from it.
What good looks like: managers arrive at launch already using the coach, and can say why it is useful. A manager who has not used the coach cannot ask reps about it in a one-on-one, and that question is what keeps reps using it.
6. Publish what managers can see (before go-live)
Write down, in one paragraph, what managers will see in full (roleplay transcripts, call transcripts and scores) and what they will see only as a summary (what reps are working on and where deals are stuck, not the transcript of a prep session). Send it to every rep before their first session.
What good looks like: reps bring their difficult deals. If reps only bring deals that are going well, they believe they are being watched, and the coaching stops being useful.
7. Pilot with one team on live deals (weeks four to six)
Each rep's first session is a deal of their choice with a buyer call booked in the next few days, prepared with the coach on a call while the manager listens. From then on, prep before important calls and a debrief afterward. Practice scenarios can come later, for specific needs such as a product launch.
What good looks like: by week six, most reps prepare their most important call each week with the coach. The week-by-week habit is in how to get reps to use AI coaching, and the prep method in how to prepare a call with an AI coach.
8. Review at week six and expand team by team
At week six the owner reports the measure from step one, the usage pattern and the friction reps reported. Fix the friction, then start the next team at step five.
What good looks like: a short report the sponsor reads, with the measure moving and a list of fixes made. Expanding one team at a time lets each new team's managers learn from the last.
Worked example: two teams in six weeks (illustrative)
This example is illustrative. The company, teams and timeline are invented to show the plan.
A software company with four sales teams sells with MEDDPICC. The VP of Sales has one complaint: deals reach procurement with no economic buyer named.
Week one. The head of enablement owns the rollout. The measure is the share of deals at proposal stage with a named economic buyer. Security and legal receive the vendor's documentation on day two.
Week two. The coach is configured to MEDDPICC with the company's discovery guide and product sheets. Enablement pulls 30 coaching cases from the last two quarters, and two senior managers judge the coach's answers blind against the current approach (managers' written notes). They flag one weakness, that the coach is too quick to suggest a pilot. The vendor adjusts and they re-run.
Week three. Team A's manager runs two pipeline reviews from Bryce's read of the team. She finds three deals where her reps said "champion engaged" and the read shows no buyer evidence of it.
Weeks four to six. Team A's eight reps go live. Each starts with a deal of their choice. By week six, seven of eight prepare their most important call each week. Two reps raise friction: the coach does not know a new pricing tier, and one rep's calls start too early for morning prep. Both are fixed.
Week six. The owner reports the measure moving in Team A and starts Team B at step five.
Where it goes wrong
- Launching as training. A kickoff, a scenario library and a completion target produce a busy week one and an empty week four.
- Managers last. Reps take their cue from the one-on-one. If the manager never mentions the coach, reps conclude it does not matter.
- Skipping the test. A coach that gives advice your experts would not give loses reps' trust quickly, and it rarely gets a second chance.
- A generic method. Advice that has to be translated into the team's terms does not get used.
- Surprise reviewers. Security or legal hearing about the rollout in week five can stop it for a quarter.
Rolling out across every team
The plan above works for one team with any capable coach. Across many teams, three parts of it get harder: keeping the coach on the team's method as it changes, keeping its quality tested as models change, and giving every manager a read across their deals.
Our AI sales experts are built around those three. Each runs the team's own method on live deals (Caty for MEDDPICC or a house method, Keenan for Gap Selling, Val for the ValueSelling Framework®), the rep talks to it on a call, and it remembers the deals the rep has brought. They are held to cases and criteria set by methodology experts and re-tested whenever a model is released or retired, and each manager gets a team read from Bryce (see how it works). The arguments behind the category are in the AI sales coaching guide and does AI replace sales managers.
If you have the step-one sentence and number written down, that is enough to start a first team: see for teams or talk to us.
FAQ
How do I implement AI sales coaching? Name the pipeline problem and a deal-level measure, appoint one owner, and brief security and legal in week one. Configure the coach to your method and test it on your own cases with your experts. Start with managers, publish what managers can see, then pilot with one team on live deals and expand team by team.
How long does an AI sales coaching rollout take? About six weeks, with setup, the manager weeks and the pilot overlapping: see the timeline above. Security and legal review can extend that if reviewers are briefed late, so send documentation in the first week.
Who should own an AI sales coaching rollout? One person, usually the head of enablement or a sales leader, from approval to the second team. They sequence the reviewers, set the measure, run the expert test and report at week six. Front-line managers own adoption on their teams.
Should we start AI coaching with new hires or the whole team? Start with one whole team, including tenured reps, on live deals. Tenured reps carry the deals that decide the quarter and set the example for others. New hires can join through the same deal-based habit, with practice scenarios added for the pitch and product.
How do we measure whether AI coaching is working? Measure something on the deal. Good choices are the share of agreed next moves that happen and the share of late-stage deals that have the evidence your method requires. Track usage too, but treat logins and completions as activity, since they rise and fall with novelty.
What if usage drops after the pilot? Look at the one-on-ones first. The plan above rests on two things: managers asking about the coach, and reps trusting what managers can see. Check that each manager still opens their read before pipeline review and asks reps about the deal they prepared, and resend the note on what managers can see. Then ask two or three reps what got in the way, and fix that.
Do reps need training to use an AI sales coach? Very little. The first session should be the rep's own deal, prepared with the coach on a call while the manager listens. The rep learns by doing it. A short note on what managers can see, sent before go-live, matters more than a training module.