Count The Handoffs

THE AI AUDIT

There's a number that predicts how frustrating your business is to run, and almost nobody tracks it.

It isn't revenue. It isn't headcount. It isn't how many tools you pay for, though that one's fun to be horrified by. It's how many times a single piece of work changes hands before it's finished.

Call it the handoff count. Every time a job moves from one person to another, from one tool to another, or from one inbox to a different inbox, that's a handoff. Each one is a small tax. Information gets lost. Somebody waits. Context has to be rebuilt from scratch by whoever picks it up next.

The tax on any single handoff is tiny, which is exactly why nobody notices. The tax on forty of them a week is your whole margin.

I've run this audit with a lot of businesses now, and the pattern holds with irritating consistency. The owner is convinced they have a productivity problem, or a people problem, or a "we need better project management" problem. What they actually have is a handoff problem wearing three different disguises.

So let's count them. This takes about ninety minutes and it's the highest leverage thing you'll do this week.

Here's the thing about a handoff that makes it so expensive.

When work sits in one person's hands, the context lives in their head. They know why the client wanted the blue version. They remember that the invoice was disputed last quarter. They know the deadline is soft but the person asking is not.

The second that work moves, all of that context has to be either written down, which nobody has time for, or reconstructed, which is slow and lossy, or skipped, which is where mistakes are born.

That's the real cost. Not the two minutes it takes to forward an email. The twenty minutes somebody spends the next day figuring out what they're actually looking at, plus the follow up message asking a question that was already answered somewhere in a thread nobody can find.

And handoffs compound in a way that feels deeply unfair. Two handoffs isn't twice the friction of one. It's more, because by the third pair of hands, the original intent has been through a game of telephone and somebody is now solving a slightly different problem than the one that came in.

This is also why hiring often fails to fix the feeling of being overwhelmed. You add a person, and you think you've added capacity. What you've often done is added handoffs. The work still takes the same total effort, it just now has more seams in it, and seams are where things tear.

Pick one week. Last week is ideal because it's fresh and you can still reconstruct it.

Pick three things your business does repeatedly. Not the exotic stuff. The bread and butter: a new lead comes in, a project gets delivered, an invoice gets paid. Whatever your three most common flows are.

Now trace each one, start to finish, and write down every single time the work changes hands.

Be brutal about what counts. A handoff isn't just person to person. All of these count:

Somebody moves information from one tool to another, even if it's the same person doing it. Copying a lead's details from the contact form email into the CRM is a handoff.

Somebody has to ask somebody else for something before they can continue. That's a handoff plus a wait.

Work moves from a shared space into a personal one, or the reverse. A file that lives in someone's downloads folder for two days is a handoff that hasn't landed yet.

Something gets re-entered. Anywhere the same fact gets typed twice, you've found a handoff.

Write them in a list, in order, with a name next to each one. Who touched it, what they did, what it was waiting on.

Most people are genuinely surprised by the count. A process the owner describes as "pretty simple, we just take the order and fulfill it" turns out to have eleven handoffs, four of which are the same person moving data between two systems that don't talk.

Now the useful part. Not every handoff is bad. Some are load bearing. The goal isn't zero, it's honest.

Pile one: handoffs that add judgment. Somebody looks at the work and makes a decision that requires knowing something. A quote gets reviewed before it goes out. A design gets a second opinion. These are earning their cost. Leave them alone. In fact, protect them, because these are usually the first thing people try to automate away and they're the ones that shouldn't be.

Pile two: handoffs that only move data. Nobody decides anything. Information goes from one place to another place because the two places don't talk. This is the pile that should be nearly empty and never is. Every single one of these is a candidate for automation, and the good news is that this is the easiest kind of automation there is, because there's no judgment to encode. You're just building a pipe.

Pile three: handoffs that exist because of an old decision. Somebody has to approve something that hasn't needed approving in two years. Work routes through a person who used to own that area and doesn't anymore. A step exists because of one bad incident in 2023 and nobody has revisited it. These are the most satisfying to kill and the most political, because there's usually a person attached who thinks that step is their job.

Do the sort before you touch anything. The instinct is to start fixing the first thing you find, and the first thing you find is almost never the worst thing.

THE AI WORKFLOW BLUEPRINT • $47

The Blueprint is the twelve workflows I map for every business I audit, with the handoff points already marked and the automations specified. Instead of starting from a blank page, you start from a diagram of what good looks like and adjust it to your business.

Pile two is where the money is, so let's be specific.

For each data only handoff, you're asking one question: what would it take for this information to arrive where it needs to be without a human carrying it?

Sometimes the answer is a native integration you already have and never turned on. Genuinely, check this first before you build anything. A depressing percentage of the manual copying I find in businesses exists because nobody clicked through the settings page of a tool they already pay for.

Sometimes the answer is a connector. This is what Make is for, and it's the single highest return tool in most small business stacks precisely because it's boring. It doesn't do anything clever. It moves things from A to B reliably, on a trigger, forever, and it doesn't get bored or go on holiday.

Sometimes the answer is that the two systems shouldn't both exist. If you're moving data between two tools constantly, and neither one is doing something the other can't, you've found a consolidation opportunity rather than an automation one. Automating a pipe between two redundant tools is how you end up maintaining a pipe between two redundant tools.

And sometimes the honest answer is that the handoff needs a human because the data arrives in an unpredictable shape. That's where AI actually earns its place, and it's a narrower slice than the hype suggests. Reading a messy inbound email and pulling out the structured details is a real job that language models are genuinely good at and rule based automation is genuinely bad at. Use it there. Don't use it to replace an integration that already works.

Start with the handoff that happens most often. Not the most annoying one, the most frequent one. Frequency beats severity for return on effort, almost every time. The thing that irritates you most is usually the thing that happens once a month and sticks in your memory because it's irritating.

There's a second number worth capturing while you're in here, and it's the one that actually explains why your customers think you're slow.

For each handoff, write down how long the work typically sits before the next person picks it up.

This is where it gets uncomfortable. In most businesses, the total time spent working on something is a small fraction of the time between the request arriving and the thing being done. A job that takes four hours of actual effort takes nine days to deliver, and eight and a half of those days are the work sitting in a queue.

Your customer doesn't experience your effort. They experience the elapsed time. Which means you can make your business feel dramatically faster without anybody working harder, purely by attacking the waiting rather than the working.

The fix for waiting is usually not automation. It's a rule. Work doesn't sit in a personal inbox. Handoffs go to a shared place with a visible state. Anything waiting more than one working day surfaces itself without somebody having to remember to chase it.

That last one is worth automating, and it's about a twenty minute build. A scheduled check that looks at anything sitting in a stage too long and posts it somewhere visible. Not an email, because emails become noise. Somewhere the team already looks.

Since this is nominally an AI newsletter, let me be direct about where the technology actually fits in this audit, because the honest answer is narrower than most people selling it will tell you.

AI is good at the handoffs where information arrives unstructured and needs to become structured. Inbound emails, call notes, form submissions with a free text field, PDFs from suppliers who will never adopt your preferred format. Pulling clean data out of mess is the job, and it does it well enough now that it's silly to have a person doing it.

Tools like Fathom are a good example of what this looks like when it's done right. The handoff used to be: have a call, take notes badly, remember to write up the notes, remember to put the action items somewhere. That's three handoffs and two of them regularly failed. Now the call produces the summary and the actions on its own, and the handoff is one step instead of four.

AI is bad, or at least unnecessary, at the handoffs where the data is already structured and the rule is already clear. If a paid invoice needs to update a status field, you don't need a language model. You need a webhook. Using AI there adds cost, latency and a small but real chance of it doing something imaginative.

The audit tells you which is which. That's the whole point of doing it before you buy anything.

Trace three flows. Count the handoffs. Sort into judgment, data, and legacy.

Kill two things from the legacy pile this week. They're free wins and killing them builds the political capital for the harder changes.

Automate the single most frequent data handoff. One. Not five. Get it working, watch it for a week, then do the next one.

Add elapsed time to your three flows and find the longest wait. Fix the rule, not the tool.

Then run the count again in a month. The number should be meaningfully smaller, and if it isn't, you automated the wrong things, which is useful to know while it's still cheap to find out.

Most businesses don't need to work harder or buy anything new. They need to stop dropping the baton eleven times on the way to the finish line.

THE AI BUSINESS ACCELERATOR • $97

Eight weeks of doing this properly with your actual business. We map your flows, count your handoffs, build the automations in priority order, and document them so the whole thing keeps running after you stop paying attention to it. You leave with a system, not a subscription.

Jordan

The AI Newsroom | Jordan Hale | ainewsroomdaily.com