Every trial carries something, so “risk-free” is a phrase worth a second look. The truth underneath it is narrower and more useful. Your mail is safe here, because a new email client signs in to the mailbox you already have. One thing in a trial genuinely can go wrong, and it has a setting. The bigger exposure is two weeks of the firm’s attention spent on an evaluation that ends in a shrug. Give the window a question to answer and most of the rest looks after itself.
- Your mail stays where it is. A new email client signs in to the mailbox you already have, so there’s nothing to move either way.
- You end the connection from your own provider’s settings. Your account is the only side that has to cooperate, and that’s the real safety net. It’s worth confirming before you connect rather than after.
- One operational risk is genuinely live. Something goes out in your name before you’ve watched the software long enough to know how it writes. It has a setting, and the setting starts in the right place.
- The larger risk is a trial that settles nothing. Run it too cautiously, at the wrong time of year, with the question in your head instead of on paper, and you end up with a feeling where a decision should be. Then you pay for the whole exercise again in eight months.
- Your existing obligations already cover connecting real client mail to a new supplier. A trial sits inside them. It’s the same disclosure, made earlier.
Three different things get called risk
Ask an owner what they mean by trying something without risk and you’ll usually get three things at once. The three have wildly different sizes. Separating them is most of the work.
Risk to the asset. The mail, the history, the address, the contacts. This is what people picture, and in this kind of switch it’s close to nothing, because everything stays where it is. Google or Microsoft holds your mailbox. An email client is an app you point at it. So there’s no export, no import, and no far end that has to end up matching the near one. Why there is no migration when you switch email apps works through why that’s structurally true rather than a promise anybody is making.
Risk inside the window. The two or three weeks while you’re still learning the software. A message goes out that reads slightly wrong. Something important sits a row lower than it should have. It’s real, it’s small, and every part of it has a specific, cheap bound. Most of this page is about those bounds.
Risk to the decision. The hours, the attention, and the fact that an abandoned switch costs nearly the full price of a completed one and leaves you where you started. It’s the biggest number by a distance, and the one firms leave out of the plan. What switching email tools really costs prices the whole exercise for a small firm, and the largest line in it is the deciding rather than the doing.
That ordering is why the asset gets two paragraphs here and the decision gets most of the page. The expensive outcome is a trial that ends without one.
A trial with a question in it
Most trials carry an intention rather than a question. The intention is to have a look, and two weeks of looking gives you an impression. Then the impression fades, the free period lapses, and four months later you’re reading the same three comparison pages again, with no memory of what you concluded.
So write the question down before you connect anything. Make it specific enough to be wrong about. “Is this any good” drifts, and these land:
- Does client mail still end up below the fold on a Thursday afternoon in busy season?
- Does the thing I’m waiting on get chased without me remembering it?
- Can I read a forty-message thread as a line, accurately enough that I stop opening it?
- Do I get through the first pass of the morning in materially less time than I do now?
Take two or three, from whatever actually made you start looking. Write them where you’ll find them again. Write the failure condition next to each one, because a criterion you can only pass is decoration. “If I am still opening every thread by the end of week two, the answer is no” ends a trial cleanly. “It seemed fine” leaves it open.
One more thing to settle before you connect: run it on real mail. An empty spare mailbox gives the software no history to rank, no relationships to learn and no pressure to handle, so you’d be testing whether it runs rather than whether it helps. The careful version of that instinct is to leave the settings where they arrive, which the next two sections are about. What to check before you connect your inbox covers the checks that genuinely do belong before the first sign-in.
The window that cannot cost you anything
Four decisions make a trial reversible in practice rather than in principle. Each one takes about a minute. Make them in this order.
Keep the old app installed and signed in. This is the whole safety net, and it’s free. For two weeks you’re running two readers over one mailbox, and either of them can be the one you open on a given morning. Notifications, though, belong on exactly one of them. Two apps alerting on the same message is the single most common reason a trial gets abandoned in week one, and the new thing takes the blame for it.
Change nothing else at the same time. Leave the filters, leave the labels, leave the folders alone. There’s a cost argument for this, and the switching cost page makes it: one new variable is diagnosable and four are not. In a trial there’s a second reason. Everything you change is something you’d have to change back. Leave the old setup intact and backing out is closing a tab.
Keep it invisible from the outside. Hold the new booking link, keep your old signature where it is, and let clients carry on as they are. Your address stays the same, so everyone outside the firm already has what they need, as why you don’t need a new email address sets out. Here the reason is narrower. A trial the outside world can’t see is one you can end quietly.
Use one mailbox, and make it yours. Your own, rather than the shared info@, for as long as you’re evaluating. A mistake in a shared mailbox is seen by the most people and owned by the fewest. Shared access also carries the most awkward setup questions, and those are better asked once you’ve decided. If you run a second, quieter business, its mailbox is a reasonable first choice, because the mail in it is real.
Where the dial belongs while you are still deciding
Here’s the part that runs against the instinct. It’s the most common way a careful firm spends two weeks and still has to guess.
On day one the instinct is to set everything to the most cautious position available and watch from behind glass. In a product with an autonomy setting, that means suggest only. The software works out what it would do and shows you, and the doing stays with you. It feels like the responsible choice. It’s also the one setting that guarantees your question goes unanswered, because at suggest only you’re still doing all of the work. Two weeks later the honest finding is that it saved you no time, and that finding is about the setting rather than the software.
The evidence lives at the middle position, where the work gets done and nothing has committed. The reply is written. The meeting time is found. The buried ask is a dated task. All of it waits on your yes. That’s where a trial shows you both things at once: whether the judgment is any good, and whether the finished work is worth having. Setting how much your inbox does on its own explains why the two lower positions differ in how much of the job gets done rather than in how much you decide, and it’s worth the five minutes before you start.
Two rules for the window itself.
Do not raise anything to fully handled during the trial. The reason is evidence rather than recklessness. Fully handled removes the step where you’d have seen a bad one, and seeing the bad ones is the entire purpose of the two weeks. Raising a setting is a decision you make with evidence you’re still collecting. It’s the reward for a trial that went well.
Leave replies where they start, whatever else you do. This is the one line on the page with a genuinely irreversible failure on the other side of it. Filing is reversible. Ranking is reversible. A booked meeting moves with one short message. A message that reached somebody’s server sits on a system you don’t control, which undoing what Point did is honest about. Everything else in a trial can be put back.
If the thing you want to test is whether the judgment can be trusted, watch the work at review and read the record. Raising a setting answers a different question. AI email triage, and whether you can trust it sets out what watching looks like when it’s done properly, and the activity log behind every action is where the evidence gathers while you’re busy.
What a trial does not suspend
Three obligations attach on the day you connect rather than the day you sign an annual contract. All three are light. All three are cheaper to deal with while you’re still a prospect, because that’s when your questions get answered fastest.
Your duty to pick and oversee a supplier. A firm inside the FTC Safeguards Rule picks a service provider the moment it points a new app at client mail. The FTC’s own business guidance names “tax preparation firms” among the examples of covered financial institutions (FTC Safeguards Rule: What Your Business Needs to Know, checked September 7, 2026). The rule itself requires you to oversee service providers by “Taking reasonable steps to select and retain service providers that are capable of maintaining appropriate safeguards for the customer information at issue” and “Requiring your service providers by contract to implement and maintain such safeguards” (16 CFR 314.4(f)). That sentence applies from the first connection, invoice or no invoice. In practice it’s a short list of documents to collect in week zero, which is what questions to ask any AI tool about your data is for. The confidentiality side of the same decision is using an AI inbox without breaking confidentiality.
The floor the platform already imposes on the app. Some of what you’d otherwise take on trust is set by the company that holds your mailbox rather than the one you’re evaluating. Google’s Workspace user data and developer policy applies to applications using sensitive and restricted scopes, and it puts training among the things it flatly prohibits. The policy’s own words are “Transferring, selling, or using user data to create, train, or improve a machine learning or artificial intelligence model beyond that specific user’s personalized model for the appropriate use case or user-facing feature” (Google Workspace user data and developer policy, checked September 7, 2026). Read that as a floor rather than as an assurance. It sets the limit on what a compliant app may do with Gmail data. A particular vendor’s retention and subprocessors are a separate set of answers, covered in keeping client data out of AI model training.
Whatever the trial’s own terms say about money. Note the conversion date the day you start, in the calendar, with a name against it. The legal floor under an online subscription that renews itself is the Restore Online Shoppers’ Confidence Act. It makes charging for anything sold online through a negative option feature unlawful unless the seller discloses the material terms before taking billing information, obtains express informed consent before charging, and “provides simple mechanisms for a consumer to stop recurring charges” (15 USC 8403). Worth knowing, because there’s a stronger rule you may have read about. The FTC’s amended Negative Option Rule, the click-to-cancel one, was vacated in full by the Eighth Circuit on July 8, 2025 (Custom Communications, Inc. v. Federal Trade Commission). So the statute is the floor that stands. The rest of the money question, including what you’re still paying for on the tool you might replace, belongs to what switching email tools really costs.
Deciding on a date you fix now
Put the decision date in the calendar before you connect. Two weeks is the right length for almost every question a trial can actually settle. Fix it in advance, because a trial fails by running on rather than by ending badly. A month in, the new thing is simply what you use, the decision never got made, and you’re paying for a tool you never evaluated.
What two weeks can genuinely produce evidence about:
- Whether the right things are at the top when you sit down, judged against mornings you remember.
- Whether a summary of a thread you already know is accurate. This is the cheapest test on the list, because you’re the answer key.
- Whether a draft arrives closer to sendable than to blank, in your words rather than in a generic register.
- Whether anything got missed. You’ll know, because the old app is still open.
- How much of your first pass disappeared. Rough is fine here. You’ll notice ten minutes, and you’ll certainly notice thirty.
Then write one sentence down. Yes because of this, or no because of that, or one more week to settle a specific thing that’s still open. The sentence is the deliverable. It’s what lets the eight-months-later version of this exercise start where you left off, and it’s worth more than everything else you did in the window. What the first productive week tends to look like from the inside is getting value in the first week, and the day-by-day version is your first two weeks with Point.
Four things two weeks cannot tell you
Here’s where “without the risk” stops being literally true, and saying so beats glossing it. A well-designed trial bounds the reversible risks and leaves four real ones standing.
How it behaves in your worst month. A trial run in June is silent about April. April is the most expensive two weeks of the year to be learning anything, so the remedy sits elsewhere. Name what April specifically needs. Check that the mechanism for it exists, and ask the vendor how it holds up under volume. That’s a weaker test than watching, and it’s the best one available.
Whether the firm will adopt it. One person’s trial is evidence about one person, and that person chose the tool and wanted it to work. The bookkeeper had it arrive on their desk. That gap is the most underestimated line in any switch, and it has its own remedies, which are in bringing your team along on a new tool.
Whether the vendor will still be here in three years. Two weeks has no way to reach this one. So ask in week zero what you’d walk out with: whether tasks, notes and settings can be exported, and what happens to your data after you cancel. A vendor should have a straight answer, and if you ever want to leave, what stays and what goes is the shape that answer should take.
Whether it keeps getting better at your work. Two weeks of corrections is the start of a curve. Most of what makes this kind of software useful in month six is built from what you fixed in month one, so a trial sees the product at its weakest. That cuts in the tool’s favor, which is exactly why it’s worth stating. Close to worth it after two weeks is probably comfortably worth it after two months. Far from it after two weeks usually stays far from it with more time.
Backing out, and what stays behind
Ending a trial is a revocation rather than a move, and you make it from your own account.
On Google, the linked apps page in your account settings lists every app with access and removes one on a click. Google states the effect plainly: “If you remove access, the app can’t access your Google Account. This may make some features unavailable” (Google Account Help, checked September 7, 2026). On Microsoft 365 the equivalent lives in the My Apps portal. In Microsoft’s words, “You can revoke any of the permissions you consented to by selecting Revoke Permissions”, with the caveat that “removing a permission may break some of the apps functionality” (Microsoft Support, checked September 7, 2026). Either way it takes under a minute, and your account is the only side that has to act.
Two things stay behind after a revocation, because “nothing is stranded” is a rosier sentence than the truth.
The copy synced while the connection was live stays, along with the index and summaries built from it. Those are separate objects on the vendor’s retention schedule. Ending access and asking for deletion are two different requests, and the second one is worth making in writing. Where your mail goes when AI reads it sets out what those objects are.
Whatever the app wrote into your own mailbox stays too, which mostly you’d want. Labels or categories it added, mail it filed, drafts you sent, calendar entries it created: those are ordinary items in your account now, the same as ones you made yourself. What goes is the layer that was doing the thinking, and the corrections that taught it. That trade is the honest cost of stopping, and it’s set out properly in if you ever want to leave.
A two-week trial, in order
- Write the two or three questions down, with their failure conditions. Ten minutes, before anything is connected. It’s the step that gets skipped, and the one that decides whether the two weeks were worth anything.
- Do the pre-connection checks. Whose decision this is, what’s already connected, what’s in the mailbox that needs a document first. The full list is here, and it’s twenty minutes.
- Collect the vendor’s documents while you’re a prospect. Retention, subprocessors, what you leave with, what the trial converts to and when.
- Pick a week where a mistake is cheap. Clear of the first week of April, clear of the run-up to October 15, and with everybody in.
- Connect one real mailbox on a quiet afternoon, and leave every setting exactly where it arrives.
- Send yourself the first message. From line, signature, anything appended at the bottom. Thirty seconds, and it settles the only thing a client could notice.
- Turn notifications off in the old app, and leave the app itself installed. One source of alerts, two ways to read.
- Change nothing else for two weeks. No pruning, no tidying, no new booking links, no announcements.
- Read the activity record twice before the date, as well as the feed. The feed shows you the good days. The record shows you everything.
- On the date you fixed, write the sentence. Yes, no, or one more week for one named reason. Then either close the old app or revoke the grant, and in both cases do it that week rather than letting it drift.
What a trial of Point looks like
Point is an email client, so a trial starts with the Gmail or Microsoft 365 sign-in you already use. Your address stays. Your mail stays where it is. A couple of clicks later Point is reading the mailbox you already had. That’s how it works after a trial too, which is the point: there’s a single setup, and no evaluation mode to get out of afterwards.
The setting that matters most during an evaluation already sits where this page argues it should. Point holds a separate autonomy setting for each kind of work rather than one switch across all of it. Every one of them arrives on review, the position where the work gets done and nothing has committed. A safe trial is what you get out of the box, and a useful one is what you keep by leaving it there.
While you watch, Point keeps a record for you. Every action Point took is listed there in plain language with the time against it. You reverse one from that list, so a thread Point filed is back in your feed as soon as you disagree with the filing. Reversal reaches Point’s own side of the line and stops there, which is why replies are the setting to leave alone while you’re still forming a view.
Two things make a Point trial unusually visible from inside the setup you already have. On Gmail, three labels appear, Now, Later and Other, and every triaged message carries the one it was given, so the sorting reads from the real mailbox as well as from the new app. On Outlook the same three show up as categories. Filing a thread in Point takes it out of the inbox on the provider’s side as well and keeps it, so it sits where the rest of your archived mail already does. Your old app, in other words, is the fallback during the two weeks and a second view of the same evidence.
If you run more than one business, each one stays in its own lane during the trial as it would afterwards: every card names the business it belongs to, and one click narrows the whole app to just that one. Point is in private beta and its terms are per seat, so what a trial costs and how long it runs are questions to put to the team rather than assumptions to make from this page. Everything Point does is the full inventory if you want to write your questions against it.
Common questions
How long should a trial of an email client be?
Two weeks, fixed in advance, for almost every question that a trial can actually answer. A shorter window leaves you still finding the buttons. A longer one lets the trial quietly become the status quo while the decision goes unmade, which is the failure mode to design against rather than the cautious option. If one specific question is still open on the date, extend by a week for that named question and no other.
Can I run the trial on our shared mailbox?
Use one person’s real mailbox instead. A mistake in a shared mailbox is seen by the most people and owned by the fewest, and shared access brings its own setup questions, which are easier to answer once you’ve decided than while you’re deciding. If the answer is yes, the shared mailbox joins the rollout rather than the evaluation.
Is it safe to point a trial at real client email?
It’s the same question as pointing anything at real client email, and it has the same answer. The supplier’s documents decide it, whether you’re paying yet or not. Collect them in week zero. For a firm inside the FTC Safeguards Rule, the obligation to select a capable service provider and put safeguards in the contract attaches when the connection is made. A trial is a good reason to have that conversation early.
Do I need to tell clients I am trying a new email tool?
Clients see exactly what they saw last week, because your address stays the same and threads keep threading. Whether you disclose the use of AI tooling in your engagement terms is a separate question with its own answer, and it stands whether you’re in a trial or out of one. Settle it before you connect rather than after.
What if I decide against it halfway through?
Stop that day rather than serving out the two weeks. Revoke the grant from your own provider’s settings. Open the old app that’s been sitting there the whole time. Write the reason down in a sentence, so the next person who raises it in a partners’ meeting gets your two weeks for free. An early no is the cheapest outcome available, and it counts as a trial that worked.
Does starting a trial commit me to a subscription?
The terms you agreed to decide that, which is why the conversion date belongs in the calendar on day one. Under ROSCA, an online subscription with a renewal feature has to disclose its material terms before taking your billing details, get express informed consent before charging, and give you a simple way to stop recurring charges. The broader click-to-cancel rule was vacated in 2025, so that statute and your state’s auto-renewal law are the floor, and the vendor’s own terms are the specifics.
The short version
- Separate the three risks. Your mail stays where it is, the window is bounded and cheap, and the expensive one is an evaluation that ends without a decision.
- Write two or three questions and their failure conditions before you connect. A written question is what turns two weeks into an answer.
- Keep the old app installed with its notifications off, change nothing else, keep the trial invisible to clients, and use one real mailbox that’s yours.
- Leave the autonomy settings where they arrive. Suggest only guarantees a null result, and fully handled removes the step where you’d have caught a bad one.
- Collect the supplier documents in week zero. For a firm inside the FTC Safeguards Rule, connecting a mailbox is choosing a service provider, trial or not.
- Fix the decision date now and write one sentence on it. Trials fail by running on.
- Be honest about the four things two weeks can’t show you: your worst month, your team, the vendor’s future, and how good it gets after you’ve taught it.
- Backing out is a revocation you make from your own account. Access ends immediately, and what was copied while it ran is a separate question worth asking in writing.