The handoff sheet
Why AI made your people faster but not your organization, where the time actually goes, and a one-page sheet for the first handoff you fix.
Reading time: ~7 min | Issue #3 | Book a Discovery Call
Happy Wednesday, Brady here.
On October 15, Vertafore started selling two agents for the least glamorous work in an insurance agency: checking a policy against the agency's records and flagging what doesn't match, for a person to approve. Somebody is now selling, as software, the checking people do by hand when work moves from one desk to the next.
The same week, we were on a discovery call with a retailer whose sales leader walked us through roughly eleven forms between a proposal and an invoice, with the same information re-keyed into each one. Some version of those eleven forms exists in your organization too, and that gap is what this issue is about.
Inside the issue
Why AI made your people faster but not the business, and a week-by-week way to start on one handoff
Worth Your Time: the handoff sheet, six questions to answer before anyone builds anything
Field Notes: the hours companies now spend checking agents, the handoffs that are just a person moving work between tools, and Microsoft's agent accounts
Your people aren't slow. The handoffs between them are.
In McKinsey's latest global survey, 80% of respondents said AI had made them more productive, but only around 35% could point to any profit impact from it. If you run an organization, you've probably felt that gap without needing the numbers: your people use AI every day now, and the organization doesn't seem to move much faster for it.
Most organizations we work with use AI to make individual steps faster, a better first draft of the proposal, a quicker summary of the call. That's real, and it's where everyone starts. But the bigger opportunity sits between the steps, at the point where work passes from one person or team to the next and waits there while somebody checks it by hand before they trust it enough to begin. McKinsey calls the cost of those human handoffs the coordination tax, and at one large manufacturer they measured it: the work inside the steps of a planning process took 12 to 24 hours, and the waiting between the steps took 9 to 18 days.
We've seen this in the last couple of weeks at an association, a professional services firm and a hospitality company. How they operate doesn't have much in common, but the cost was the same in all three: capable people waiting on other capable people to look at something before the work could move.
Earlier generations of automation mostly left those handoffs alone, because they couldn't make the judgment call where one person's work becomes another's, like whether the order matches the contract. Agentic AI, software that reads what comes across and checks it against what should be there, can now do that checking and hand a person only the mismatches. Last week, Vertafore started selling exactly that to insurance agencies. At the companies we work with, the waiting at those handoffs is the one cost nobody has a budget line for.
How to start on one handoff
Cutting the coordination tax means rethinking work at the handoffs. It consists of six steps with the first four taking place in the first week, before anyone builds anything.
To make it concrete, take an automation for a professional services firm that runs the handoff from a closed deal through to project setup. Before, the delivery lead used the contract and whatever the seller remembered to share to build the budget, the kickoff materials and the team by hand in their project tool.
Monday: ask where work waits, not where it's slow. Don't start from the process map. Send three or four frontline people one question: "Where does your work sit waiting on someone else?"
Tuesday: pick one handoff using four tests. It's the right one if:
It happens weekly or more. A handoff that happens a few times a year won't produce enough cases to learn from.
You can name both systems. Work leaves one tool or team and arrives in another, like the CRM and the project tool, or the membership system and the accounting system.
You can name both owners. One person on each side, by role.
It's close to money or clients. McKinsey's version is to start where frequent coordination and moderate judgment meet work central to the business. For a professional services firm, that usually means closed deal to project setup, or time entry to invoice. For an association, renewal to finance.
If two candidates pass, pick the one where you can measure the wait before and after without building anything new.
Wednesday: put three people on it.
The person who lives on each side of the handoff, not their bosses. They're the ones who know the unwritten rules.
One builder. Someone internal who's comfortable with your systems, or outside help. One person, not a committee.
You, as sponsor. Your job is to clear access, protect their time, and decide early what happens to the hours you free up, then tell the two owners before anything runs.
Thursday and Friday: write down what "right" looks like. Get the two owners in a room for an hour. What has to be true for the work to pass cleanly? What do you check today, by hand, before you accept it? That list becomes the first set of checks.
The next week: build the checker, and agree when you'd turn it off. The builder sets it up to read what came across, run the list and send both owners a short note for every mismatch. At the professional services firm, that was a workflow that runs on its own when a deal is marked closed in the CRM, with an AI step that reads the contract, the call notes and the email thread, compares them to the project setup, and messages the delivery lead and the seller together. A person decides what to do with each note, and if most of them are noise after two weeks, or almost nothing fires, you turn it off.
Two weeks later: sit the same two people down with the pile of notes. Now the conversation is about real cases. This one fired because the customer wrote from a different email. This one fired because the seller promised a date on a call. Ask what the rule should be, and write it down. At the professional services firm, the notes that mattered were things the client heard on a sales call that never reached the contract, and the delivery lead started getting the contract right before kickoff instead of reinterpreting it later. That's the redesign, and you couldn't have had the conversation on Monday because nobody knew what the rule was yet.
On three of our engagements this month, at a consumer business, a professional services firm and a hospitality company, a rough first version put on real work turned up rules within days that only lived in someone's head, and nobody had to spend a quarter mapping the process to find them.
The companies that cut that tax out of their handoffs will run on a cost structure their competitors can't easily match, and the lead grows over time, because a handoff with a checker on it gets better every time work passes through it. Every case it flags and a person resolves becomes a rule, so the next one moves faster and lands more accurately than the last.
So pick the handoff your team complains about most, put the two people who live on either side of it in a room this week, and ask the one on the receiving end what they check before they trust what landed.
If this is your team: there's a handoff everyone complains about, the person on the receiving end can tell you what they check by hand, and nobody has ever written it down.
The Handoff Sheet
Answer these in one meeting with the two people on either side of the handoff, before anyone builds anything.
Which handoff? One that happens every week and that someone already complains about. Monthly won't give you enough cases to learn from.
What does the receiver check by hand? Ask what they look at before they trust what landed, and write it down in their words. That list is your first set of rules.
Who hears about a mismatch? Both sides. A note that only reaches the receiver is a complaint with extra steps.
What is the checker allowed to do? Read, compare and send a note. It never changes a record and never contacts a customer.
When do you turn it off? Decide now. If most notes are noise after two weeks, or almost nothing fires, you stop.
When do you review the notes? Put it on the calendar today, two weeks after the checker starts, because that's where the notes become written rules.
The test: if someone who wasn't in the room can read your answers to two and four and tell you what the checker looks at and what it can't do, you're ready to build.
01. A third of companies say staff now spend six or more hours a week checking what agents did. That's from G2's survey of 1,019 B2B decision-makers, a software marketplace's own research, and it's the receiver's by-hand check moving to a new handoff instead of going away.
02. Nearly half of all handoffs are a person carrying work from one tool to another. Automation vendor Ace Workflow timed 187 processes and found 48.9% of handoffs work that way, so the retailer's eleven forms are the typical case.
03. The best AI agent finishes about one real job task in four. In the latest revision of UC Berkeley's Agents' Last Exam, the top system passed 24% of real professional assignments, which is the case for a checker that flags and a person who decides.
04. Microsoft agents now get their own accounts, with one memory shared across everyone they work with. Microsoft's own FAQ says users can't opt out and that files shared with the agent can surface in answers to people without access to them, so decide what your checker may read before a vendor default decides for you.
One ask this week: if there's a process in your organization that feels broken, reply and tell me about it. I answer these myself.
Until next week,
Brady Fox - Partner Dual Logic
