Last-mile deal coaching

Deal risk scoring tells you which opportunities look like they are slipping. Last-mile coaching does a different job. It opens one late-stage deal that is already on the commit list, or about to be, and inspects it until there is a single next human action, with a cite the AE and the manager can both open.

A flag without an inspect is a color on a dashboard. An inspect without a next action is a recap of what the AE already typed in Salesforce. The quality bar is the inspect: a named risk, a named person, a named artifact, and one next step. Scoring can feed the queue. It does not finish the job.

Aviso, Clari Copilot, Gong, and Salesforce belong to the same class here. They are not a ranking. Salesforce holds stage, amount, close date, and contact roles. Gong holds call and email evidence. Aviso and Clari Copilot can surface deal-specific risk and a recommended next step from those sources. None of them should be cited for a close-rate lift, and none of them should move commit on their own.

If the card only says "at risk" or "engagement down," you are still in scoring. Coaching starts when the card says what to do this week, and which record proves it.

The output is one next action, not a punch list. After the forecast line, the AE and the manager should know who does what by when, and which artifact proves the current risk.

A cite is a specific record, not a feeling about the account. "Legal has it" is not a cite unless you can name the reviewer and the last artifact. Useful outputs look like this:

• Missing signature. Next action: the AE pings the named signer about the outstanding envelope. Cite: envelope sent from the opportunity on 12 March, never viewed, no countersignature. The risk is not "legal." The risk is a document nobody has opened.
• Stalled legal. Next action: the AE or your counsel asks the buyer's named reviewer for status on the open redlines. Cite: last outbound markup dated 4 March, nothing inbound since, and contract redline AI still showing the same three open clauses. Silence is not a lost deal. It is an unanswered thread.
• Missing economic buyer. Next action: the AE asks the champion for a named intro to whoever can release budget or sign. Cite: buying-committee mapping still has an empty buyer seat; the 18 February call has the champion saying they "need finance," and finance has never been on a meeting.

Bad output is generic coaching with a confident tone. "Multi-thread the account" is not an action. It names no person, no meeting, and no cite. "Create urgency" and "confirm the close plan" fail the same test. If the model cannot point at a field, a recording, a thread, or an envelope, it does not yet have a recommendation.

Keep the card to one action even when the deal has three problems. Rank the risks, pick the one that blocks a signature in this period, and park the rest. A manager cannot inspect five next steps in a two-minute forecast line.

Coaching is evidence for the commit conversation. It is not a write-back to the forecast.

The AE confirms or rejects the recommended action. The manager then keeps or changes commit. If the card is wrong, they say so, with a reason, and the number stays a human call. Auto-changing commit from a coaching score is how this use case dies. Reps will manage the card instead of the customer.

Use the inspect to argue the line, not to replace it. A deal can stay in commit with a named stall if the manager believes the next action will clear it this period. A deal can leave commit even if the model is quiet, if the manager does not believe the signature path. Probabilistic forecast roll-up can show a range. It still does not get to relabel the AE's commit without the manager.

Do not sell this as a close-rate program. You are trying to make the last mile inspectable: one action, one cite, a confirmed commit. Whether win rate later moves is a measured-quarter question, not a claim to print on the card.

Restrict the queue to deals in commit or best-case for the current period, above the dollar threshold you already use in forecast. Running last-mile coaching on every mid-funnel opportunity is pipeline review with extra words.

For each deal on that list, require four fields on the card before the call:

1. Named risk. Missing signature, stalled legal, missing economic buyer, or another risk you can point at. Not "slipping."
2. Single next action. A human does one thing. Name the person on your side and, when you know it, the person on theirs.
3. Cite. Opportunity activity, a Gong call or email, a Salesforce contact role, or contract status.
4. AE confirmation. The AE accepts the action, replaces it, or says the cite is stale.

In the forecast call, keep a fixed order. The AE speaks first: accept the action, replace it with a better one plus a cite, or mark the cite stale. The manager does not talk commit until that is settled. Then the manager says stay, slip date, or pull. Action first, commit second, is the inspect. Commit first, then "and also multi-thread," is the old call.

If legal has gone quiet, the default next action is get a status from a named reviewer, not pull the deal. Quiet legal is ordinary on enterprise paper. Treating every silent general counsel as a lost deal will yank good opportunities out of commit and train AEs to hide legal.

If the recommended action is "multi-thread," send it back. Demand a named intro, a named meeting, or an honest empty seat on the committee. If the model wants to move commit because a score dipped, ignore the write-back. Log the inspect, keep or change commit by hand, and go to the next line.

After the period, check whether the named action happened and whether the cite was real. When the deal does sign, signed-contract data extraction is a different job: pull ARR, term, and renewal into Salesforce. That is handoff hygiene, not last-mile coaching.

The following is a made-up but realistic forecast week. It is not a measured program and not reported results.

An AE has a mid-market expansion in commit. The champion said "we're good" on last Thursday's Gong call. The Salesforce close date is this Friday. The AE's forecast comment is "verbal, waiting on paper."

A card that only flags risk would see healthy engagement and leave the deal alone. The inspect should not. Opportunity activity shows the MSA went to the buyer's legal team on 2 March. There is no viewed event. Gong has no call with anyone from legal. The committee map still lists the champion as the only live contact. Redlines have not moved.

A useful card would read:

• Risk: stalled legal. This is not a missing signature yet. The signer has not been asked to sign because legal has not released the paper.
• Next action: the AE emails the champion today and asks for a named legal reviewer and a status by Wednesday. If the champion cannot name anyone, the AE books fifteen minutes with the champion to get the intro, rather than sending another "just checking in."
• Cite: MSA sent 2 March, no view; no legal attendee on calls; champion-only committee.

In the forecast call the AE confirms that action. The manager leaves the deal in commit because there is a named path this week, and writes the Wednesday checkpoint on the line. They do not let the model pull commit because legal is quiet. They also do not let the AE keep "verbal" with no next step.

If Wednesday comes back empty (no named reviewer, still no view), the manager inspects again. That is when commit can change. The change is still the manager's.

The tools did not close the deal. They made the stall inspectable: one action, one cite, a human commit.