A firm adopts a tool one person at a time. Six people take it up separately, or five do and one quietly stays put. About ten days decides which. The things that decide it look too small to plan, and they’re the whole of it.
- The gap you’re managing is motivation. You read about this for two months and chose it. Everyone else meets it on a Tuesday.
- The unit of adoption is one person and one thing they already wanted. The same demonstration for everybody lands for the person who was already sold, and passes the rest by.
- Send one person a week ahead of everyone else. Pick them for their inbox rather than their enthusiasm.
- Check whether your Google or Microsoft administrator settings let people connect. That gate stops more rollout mornings than any objection does.
- Set a date to turn the old thing off, and set it on the day you start. Two ways of working inside one firm is the failure, and it stays until somebody ends it.
The one who chose it, and the five who did not
You’ve been thinking about this since June. You read the comparisons, sat through a demo, and worked out what it would cost. By the time you got to yes, it felt obvious. That process left you with a reason, and you’re the only person in the building holding one.
Everyone else meets the same software as an interruption, in a week they’d already planned. The bookkeeper’s March goes better the less they have to learn. The administrator has a way of working that took two years to get right, and it’s about to change. The two months of reading that made this feel worth it are yours, and they stay yours.
That’s the whole problem, and almost every rollout mistake comes from calling it something else. Firms read low adoption as a training gap and answer it with a longer session. They read it as a communication gap and answer it with a better email. The person doing the work already knows how, and already read the email. What they’re still waiting for is a reason that belongs to them.
Two things follow. The rollout is short. Most of the work is done in the first ten days, and one still limping along in week six has usually failed rather than slowed. The second thing is about you. You chose it, you’re motivated, and you’ll be fluent in three days. Your own three days tell you about you. The person who chose nothing has a different week, and that’s the week to plan for.
This page is about the mechanics of those ten days. For what the switch costs the firm in total, there’s what switching email tools really costs. For the evaluation that comes before any of this, there’s trying a new inbox without the risk.
Give each person one thing they wanted anyway
There’s a finding worth borrowing here, from the paper that started the academic literature on why people use software. Fred Davis, in Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology (MIS Quarterly, 1989), measured two things separately: how useful people believed a system was, and how easy they believed it was. Both predicted whether they used it. Usefulness predicted it considerably better, in both of the studies the paper reports.
That’s a small finding with a large consequence for a rollout. The instinct is to sell ease. It’s quick, it’s easy, you’ll have it in a minute. Ease is worth having. What makes somebody open an app on a Thursday when they’re behind is one specific thing they wanted before you bought anything.
So before you announce it, spend fifteen minutes writing one line per person. Write what that person has complained about in the last six months, rather than what the tool does.
- The bookkeeper who has asked the same client four times for the same statement, and would have to go looking to tell you which four. What they want is the chasing to carry on while somebody else holds the list, which is managing client document requests without the chasing.
- The administrator who loses an hour a week finding a time that suits three people, which is a meeting scheduler without the email chain.
- The accountant who reads a forty-message thread before a client call. Somewhere in it is the line that changed the scope, and they need to have seen it, which is how to catch up on a long email chain.
- The person who is fine, thank you, and happy as they are. There’s usually one. They get the smallest reason and the least attention, and that’s the honest answer. Leave them with the problems they already have.
Then, when you introduce it, each person hears their own line rather than a tour. Those fifteen minutes are the best-spent fifteen minutes of the rollout. Every other part works on logistics, and this is the part that works on why.
One caution about how you say it. Anything framed as an efficiency gain gets heard as a headcount conversation, whatever you meant by it. Somebody who sits outside the room where the budget gets decided hears that loudly, and firms consistently underrate how loudly. Say what it does for their week, and keep the arithmetic in your own head.
Who goes first, and why it should not be you
One person goes a week ahead of everyone else. It’s the most useful structural decision in a rollout, and it does two jobs. It finds the surprises while they’re cheap. It also produces the one piece of evidence the rest of the firm will believe.
Pick them for their mail rather than their attitude. The right first person has a busy, ordinary, representative inbox and a reputation for being straight with you. That’s often somebody quieter than the most enthusiastic person on the team. It’s almost never you. An owner’s inbox is the least representative one in the building, with more decisions, fewer repeated tasks, and a tolerance for fiddling that comes with owning the place.
That week is for finding the four things that only turn up in real use. Somebody’s signature has an image in it that comes through wrong. One person has a rule on their account that everybody has forgotten about. The client who sends everything from a personal address gets ranked in a way that surprises you. Each of those is worth about an hour to find, and finding them once rather than five times is most of the return.
Give them two instructions. Do your real work in it for a week, and keep a note of everything that surprised you. A list is enough. At the end you want that list, and one sentence on whether the rest of the firm should go. Make it easy for them to say no.
Somebody also has to own the rollout itself, and that’s a different job from going first. In a practice inside the FTC Safeguards Rule, the name partly exists already: the rule requires you to “Designate a qualified individual responsible for overseeing and implementing your information security program and enforcing your information security program” (16 CFR 314.4(a), checked September 7, 2026). A new application reading client mail lands on that person’s desk either way, so hand it to them properly. What else the rule asks of a small firm, including the training obligation that covers everyone, is set out in a simple AI policy a small firm will actually use.
Clearing the gate before anyone opens the app
Here’s the step that ruins more rollout mornings than every objection combined, and it’s a settings question rather than a people one. On many business email setups, the permission to connect a third-party application to a mailbox belongs to the administrator alone.
On Google Workspace, third-party app access is an administrator setting. Google’s documentation lists the options for apps you haven’t configured, and marks one of them as the default: “Allow users to access any third-party apps”. Where an administrator has chosen one of the others, Google states the effect plainly. “Users can’t sign in with Google to any third-party apps and websites until you configure those apps and sites with an access setting” (Control which apps access Google Workspace data, checked September 7, 2026). That’s a reasonable control, working as intended. It means one person adds the app before five people can sign in.
On Microsoft 365 the same gate is called consent. Microsoft’s documentation says that “If user consent is disabled, or users aren’t allowed to consent for the requested permissions, they aren’t prompted for consent”. So your staff meet a screen that quietly stops, rather than an error they can read. There are two ways through. An administrator can turn on the admin consent workflow, which “gives users a way to request admin consent for applications when they aren’t allowed to consent themselves” and shows them an Approval required window to act on. Or the administrator can settle it in advance. Microsoft’s phrasing is “To avoid the need for user consent, an administrator can grant consent for the application on behalf of all users in the organization” (Overview of user and admin consent, checked September 7, 2026).
The practical instruction is one line. A week before the firm goes, have one person try to connect, and watch what happens. If they get through, you’re clear. If they hit the gate, you found it on a quiet Tuesday rather than in front of everybody on the morning you’d planned.
Two related things belong in that same week. If your firm has an outside IT provider, tell them before the day rather than after. A help desk hearing the name for the first time is a rollout delayed by three days. And if anybody is on a mailbox somebody else administers, a client-side address or a legacy account, find that out now.
What to say when you tell the firm
The announcement is one short message and one short conversation, and its job is narrower than people expect. Persuasion already happened, in the per-person line you wrote. The announcement is there to settle uncertainty, and there are exactly four things people want to know.
What changes, and what stays the same. Almost everything outside the app stays exactly where it is, and naming the things beats saying it generally. Your address is the same. Your history is the same. Threads stay threaded, and the only people who hear about any of it are the people in the firm. Why you don’t need a new email address is the whole of that argument, and it’s worth reading before you write the message. People picture a migration, because the last email project they lived through was one.
When. A date for connecting, and a date the old way stops. Both, in the same message. A rollout that gives only the first date is the one still limping in week six.
Whether it’s optional. Answer this honestly, because a fake choice is the most expensive thing you can offer. If the firm is moving, say the firm is moving, and say you’ll help anybody who finds it hard. If it’s genuinely optional for one role, name the role and the reason. Imply a choice you’ve already made, then be disappointed by how somebody takes it, and your team learns that your questions have one right answer.
Who to ask. One name, and preferably somebody other than you. The person who went first is the right answer. Being the person everybody asks is also what keeps them using it.
Say it in three or four sentences and stop. A long rollout email reads as a project. A project is something people wait to see the outcome of, rather than something they do on Tuesday.
Timing is worth ten seconds of thought. Stay clear of the first week of April, and of the two weeks before an October 15 deadline. Start in a week when the person who went first is at their desk. Beyond that, a quiet week in June and a quiet week in September are the same week.
Nobody needs the same settings as you
The instinct is to configure the firm once and hand everybody the same thing. Settings sit better per person, because how much judgment you’re comfortable handing over depends on whose work it is and how much of it there is.
How much an AI inbox does on its own is a setting. In a well-built one it’s a separate setting for each kind of work, rather than one switch across all of them. The dial, and what each of its positions means is the full explanation. What matters for a rollout is that the positions are per person, and different people belong in different ones on day one.
A rough shape for a practice of six. The person who runs scheduling can go furthest soonest. A wrongly proposed meeting time is a one-line message to fix, and they’ll spot it the moment it lands. Filing and ranking can sit in the middle for everybody, because a thread in the wrong place goes back with a click. Anything that sends stays where it arrives, for everyone, for at least a month, and longer for anyone new. That’s about where the line falls rather than about seniority. Undo puts back what Point did. A message that reached somebody else’s server sits on the far side of that line, which undoing what Point did is direct about.
Two more settings decisions exist only because there’s more than one of you.
Notifications, once. Each person gets alerts from one app. During a parallel period, two sets of alerts is the most common reason somebody decides the new thing is noisy, and the noise is nearly always the old app still running underneath.
Who gets a seat, and what they can reach. A seat per person is the ordinary arrangement, and it repays a deliberate decision rather than a default one. That goes double for anyone part-time or seasonal. The finer question of who inside the firm sees which client’s mail is a real one with a real answer, and the answer lives in using an AI inbox without breaking confidentiality rather than on a settings screen.
The half hour that actually transfers
Skip the training session. A room of six people watching one screen looks like a rollout. By Thursday the one person who can still do it is the one who was always going to work it out.
What transfers is thirty minutes each, one at a time, on their real work. Sit next to the bookkeeper while they do the first pass they were going to do anyway. Leave the interface tour alone. Ask what they’re trying to get done, and let them do it badly with you there. Then teach exactly one habit, the one from the line you wrote for them, and stop.
One habit is the whole curriculum for week one. Six features taught on Monday is zero features remembered on Wednesday. People are paying attention. A habit needs a trigger, and a person attaches one new response to a familiar trigger at a time.
The second thing to teach is how to disagree with the software, and it’s worth more than any feature. Show them what to do when a thread lands in the wrong place, or a draft comes out in somebody else’s phrasing. Two things go into that. Putting the thread back takes a click. And the correction carries, because this kind of software gets better from being corrected, which is what when a summary learns from the corrections you make is about. A person who knows how to say no to software uses it more, because the question of what happens when it’s wrong is already settled for them.
Then check back once, about three days later, for five minutes. In passing, rather than as a meeting. One question: what has it got wrong so far. That question gets an honest answer, where “how are you finding it” gets a polite one.
Two ways of working is what kills it
Running the old app alongside the new one for a couple of weeks is sensible for one person. It’s the safety net that makes the whole thing reversible, and everyone keeps the old app installed.
Across a team, left running, it turns into something else. Two people file in one place and three in another. Somebody marks a thread handled somewhere only half the firm is looking. The administrator keeps two views of the same week, because one accountant is still in the old app. Now both pictures are suspect, and the firm decides the software made things worse. What actually happened is that half a firm changed and half stayed.
So the parallel period gets a fixed end. Decide it on the day you start, and write it in the calendar with a name against it. Two weeks is right for most firms. Past three, it’s a second way of working with its own constituency, and the safety net is gone.
At the end, do the closing in one afternoon: notifications off, sign-outs done, seats on the old tool canceled, any booking link clients still hold pointed somewhere current. Put it on a date, because this is the afternoon that otherwise waits forever. An old tool left open is how a four-week change becomes a four-month one.
One exception is allowed, and only one: a person, a reason, and a date. “Priya keeps the old app until the Kaminski audit ships on October 3.” That’s a decision. “We’ll see how it goes” is how firms end up with two of everything in March.
Reading week three honestly
Around week three you’ll want to know whether it took. The natural question, do people like it, buys you an answer worth very little. People are polite about things their employer bought. Four signals are harder to fake.
Somebody used it when it was inconvenient. The calm Tuesday tells you little. The real test is the afternoon everything went wrong. Were they still in it, or back in what their hands know under pressure. Reverting under load is the honest verdict, and it’s worth taking seriously rather than arguing with.
Somebody corrected it. A person who takes the time to fix a summary or move a thread back has started treating the tool as theirs. Corrections are the best early signal you have, better than compliments. Three weeks of corrections is three weeks of real use, and a person still at zero has usually barely started.
Somebody asked for something. Can it do this with the payroll mail. Can I make it stop doing that. A request is a person planning to keep using it.
It came up without you. Two people mention it to each other in an ordinary conversation, with your name nowhere in it. That’s the point where it’s how the firm works, rather than something the owner is doing.
The counter-signal is silence. Three weeks of quiet usually means somebody carrying on in the old app and hoping this passes. Ask them directly, once, and ask what’s gone wrong rather than how it’s going.
For what the first productive week feels like from the inside, so you know what you’re looking for, there’s getting value in the first week. The day-by-day version for one person is your first two weeks with Point.
The person who will not
There’s usually one, and the mistake is treating it as one problem when it’s three.
Cannot. Something concrete is in the way. Their mailbox is on a different setup. Their signature fails to come across, or they work mostly from a phone, or their machine runs it slowly. This reads as resistance, and twenty minutes of somebody sitting with them resolves it. Check for it first, every time. It’s the most common of the three and the cheapest to fix.
Not now. They’re in the middle of something with a deadline, and this week has to stay fast. That’s a fair position, and it earns a date rather than a debate. Name the date, put it in the calendar, and move on.
Will not. A genuine objection, worth hearing out before you answer it, because a fair proportion of the time it’s right. Sometimes it’s about client confidentiality, which deserves a real answer rather than reassurance, and the shape of that answer is questions to ask any AI tool about your data. Sometimes it’s about being watched, and that one is worth being straight about. An activity record shows what the software did. Say plainly that it’s there for that, and say it out loud if your team has read it the other way. And sometimes they built the current way of working and are being asked to throw it away. Ask them which part of it has to survive the change, then make sure it does.
If all of that lands and they still hold, you’re in a plain management decision rather than a software one. Separate what’s genuinely required from what’s preference. The firm’s mail lives in one place, and handoffs happen where both of you can see them. Beyond that, one person reading it in a different app is survivable. The thing to avoid is a permanent split down the middle of a firm of six, and you do that by dealing with the holdout in week two rather than finding it in month four.
A rollout, in order
- Write one line per person, on what each of them has actually complained about. Fifteen minutes, before you tell anybody anything.
- Name the person who owns the rollout. In a practice inside the Safeguards Rule, that’s very likely the individual already designated for the security program.
- Test the connection gate a week early, on Google Workspace or Microsoft 365, with one person. Tell your outside IT provider on the same day.
- Send one person a week ahead. Representative inbox, straight talker, somebody other than you, and free to come back with a no.
- Collect their list, and fix or answer everything on it before the firm goes.
- Announce in four sentences: what stays the same, the two dates, whether it’s optional, and one name to ask.
- Half an hour each, on their own real work. One habit taught, plus how to disagree with the software.
- Set the dial per person and per kind of work, leave anything that sends where it arrives, and give each person notifications from one app only.
- Check back on day three with one question: what has it got wrong.
- End the parallel period on the date you fixed, close the old tool that same afternoon, and allow exactly one named exception with a date on it.
- Read week three by the four signals rather than by whether people say they like it, and chase the silence.
Putting six people on Point
Point reads the Gmail or Microsoft 365 mailbox each person in the firm already has, so the per-person half of a rollout is a sign-in rather than a setup. Addresses stay as they are, and the mail stays where it is. Sign-in is a link sent to the person, so six people join and the firm gains zero new passwords.
The shape of a firm on Point is worth being exact about, because owners often expect the other thing. This isn’t a shared mailbox, and Point doesn’t offer one. Each person signs in as themselves, and their mail stays theirs. What the firm shares is one organization with several people inside it, and a place where they talk to each other that sits in the same feed as the mail. Tasks are personal, rather than one firm-wide list. That’s the deliberate arrangement, and what it feels like day to day, when two people stop asking each other who has something, is when you stop having to ask who has it.
Seats go on one at a time as people join, and come back down when they leave. That matters more during a rollout than it sounds. You bring the first person on, then the next three, and the whole firm commits when you’re ready. Pricing is per seat and the product is in private beta, so what six seats come to is a question to put to the team rather than something to read off this page.
The settings this page argues about are already per person and per kind of work. Every autonomy setting arrives on review, the position where the work is done and nothing has committed yet. So a new person joining in month four starts where the first person started, rather than inheriting whatever the firm has since turned up. Beside that sits a record of everything the software did, written so a non-technical partner can read it, with a time on each entry and a way back from the same list. That record is what lets you answer the being-watched question honestly. And approving anything new before it acts is how a firm keeps a newly added kind of work on a short leash until it’s been seen working.
Two things make a rollout easier to supervise from where you already are. Point writes its judgment into real Gmail labels, and into categories on Outlook, so what it decided shows up in the mailbox itself as well as inside the new app. And if the owner runs more than one business, each one stays in its own lane with its own people, which is keeping one business isolated from another. Everything Point does is the flat inventory, and it’s the right document to write your one line per person against.
Common questions
How do I get staff to use a new tool they did not ask for?
Give each person one thing they already wanted, rather than one demonstration everybody sits through. Write a line per person from what they’ve complained about in the last six months, and introduce it to each of them in those terms. Then teach one habit each, one at a time, on their real work. Check back on day three by asking what it’s got wrong, rather than how it’s going.
Should I make it mandatory?
Decide, then say which it is. A firm moving together is a legitimate decision, and people deal with it fine once it’s stated. The damage comes from implying a choice you’ve already made, because the first person who takes it at face value shows everyone else how your questions work. If it’s optional for a particular role, name the role and the reason.
Who should be the first person to try it?
Somebody with a busy, ordinary inbox who’s known for saying what they think, given a week ahead of everyone else. Pick past the most enthusiastic person, whose report comes back warm and thin, and past the owner, whose inbox is the least representative one in the firm. Their job is to come back with a list of surprises and a sentence on whether the rest of you should go.
How long should we run the old app alongside the new one?
Two weeks, with the end date fixed on the day you start. Everyone keeps the old app installed the whole time, with its notifications off. Past about three weeks it turns into a second way of working with people invested in it, and that’s the state you’re steering around.
What if our IT provider has to approve it first?
Then find that out a week early rather than on the morning. On Google Workspace, an administrator controls which third-party apps reach your data. On Microsoft 365, an administrator can grant consent for everybody in advance, or turn on a workflow that lets staff request approval from the sign-in screen. Have one person try the connection ahead of the firm, and tell your outside IT provider on the same day.
One person is refusing. What now?
Work out which of three things it is before you respond. Cannot is a concrete blocker, and twenty minutes of help fixes it. Not now is a fair position that earns a date rather than an argument. Will not is a real objection worth hearing out, because sometimes it’s right. The answer usually sits in confidentiality, in being watched, or in a way of working they built and are being asked to abandon. Deal with it in week two, because a split firm in month four is much harder to undo.
The short version
- You’re managing motivation. The rollout mistake that matters is answering a reason problem with more training.
- Write one line per person on what they already wanted. Usefulness moves people more than ease does, and this is the part of a rollout that works on why rather than how.
- Send one representative person a week early, with permission to come back with a no. Somebody past the keenest, and somebody other than you.
- Check the administrator gate on Google Workspace or Microsoft 365 before the firm goes. It stops more rollout mornings than any objection.
- Announce four things in four sentences: what stays the same, both dates, whether it’s optional, and one name to ask.
- Thirty minutes each on real work, one habit taught, plus how to disagree with the software. Then one question on day three.
- Settings are per person and per kind of work. Leave anything that sends where it starts, and give each person alerts from one app only.
- End the parallel period on a fixed date, with one named exception at most. Two ways of working inside one firm is the failure mode.
- Judge week three on corrections, requests, use under pressure and the tool coming up without you. Chase the silence.