Point vs Spark

Spark and Point are both full email clients, and they divide the work differently. Spark files your mail by what kind of thing it is, then asks you to name the people who matter and to vet each stranger who writes. Point works out what needs you today, and says what each thread wants before you open it.

• Spark’s own feature page, checked 18 August 2026, describes Smart Inbox as sorting new mail “into personal, newsletter and notification categories for easy processing, placing emails from real people at the top of your inbox”.
• The part that decides which real person comes first is you. Priority is a mark you apply to senders and threads, and Gatekeeper screens the first message from anyone new so that you accept or block them.
• That is a control-first design and a defensible one. Its limit is that a list of people who matter can only hold the people you already knew mattered.
• Point ranks every message on how important and how urgent it is as it lands, including the mail from senders you have never heard of, and how far Point goes without you is a level you set for each kind of action.

What does Spark actually sort?

Spark’s sorting is a filing rule you could explain to somebody, which is a genuine virtue. Its Smart Inbox page, checked 18 August 2026, says mail is sorted “into personal, newsletter and notification categories”, with mail from real people put at the top. Alongside that sit grouping by sender, pinning a message where you want it, and a home screen that gathers the day.

The question that rule answers is what kind of thing arrived. Is this a human being, a mailing list, or a machine telling you something happened? For most inboxes that single distinction removes a great deal of noise, and it removes it in a way you can predict and check. Nothing surprising happens. Nothing needs trusting.

It is worth saying plainly that for many people this is enough, and that Spark has been refining it for years across more devices than anything else in this comparison. A firm whose inbox is mostly newsletters and system notifications with real client mail scattered through it will feel the difference on the first morning.

Point’s sorting answers a different question, and it is a harder one: not what kind of thing is this, but does it need you today. Those come apart more often than they sound. A notification from your practice management system saying a client’s filing was rejected is a machine talking, and it is the most important thing in your inbox. A note from a real person confirming they got your last email is a human being, and it needs nothing at all.

Who decides what matters, you or the software?

Spark’s answer is you, deliberately, and it deserves a fair hearing before the objection.

Priority is a mark you apply. Its help pages, checked 18 August 2026, describe marking important senders or threads as priority so that new messages from them appear at the top of the inbox. Gatekeeper works the same way at the door: the first time somebody emails you, Spark screens the sender and gives you the option to accept them or block any future contact, with settings for whether that screening happens before the inbox or inside it. Nothing gets promoted that you did not promote.

For an owner who knows exactly who her twenty most important clients are and will spend an afternoon saying so, that is precise, transparent, and does what it says. It also never surprises her, which is worth more than it sounds when the software is standing between her and her clients. Naming the people who matter is a real technique and Point supports it too, which is the subject of VIP mail that never gets buried.

The objection is about the rest of the mail. A list of important people is built from the past, and the message that costs a small practice money is very often from somebody who is not on it: a referral from a client’s solicitor, the bank about an account you did not know had a problem, a new finance manager at a firm you have invoiced for years, a regulator. Under a nomination system, all four of those arrive as strangers. Under Gatekeeper’s screening they arrive as a decision you have to make about somebody you have never heard of, at the moment you have least time to make it. Why the important message slips past even when the sorting is working is the subject of why important mail slips past.

There is a second cost, and it is the one the owner of a seven-person firm feels most. A nomination list is a thing somebody has to build and then keep. It goes stale quietly, and the day it is most out of date is the day after your practice took on four new clients. Point weighs every message as it arrives on what it is asking for and when, so the mail from somebody new is judged on the same terms as the mail from somebody you named. You can still name people, and the naming makes a difference. What changes is that nothing depends on your having done it.

When does the AI actually run?

Both products have AI that can write, summarize, translate and search, so a feature list will make them look alike. The difference is when it runs, and that difference is most of the experience.

Spark’s is invoked. There is an assistant to ask, a writing helper for a message you are composing, and summaries and answers when you go and request them. It is sold as an allowance rather than as an unlimited part of the client: the free tier carries a limited preview and the paid tiers carry a monthly quota of AI work, per its pricing page, checked 18 August 2026. That is a coherent arrangement. If you use it twice a week it costs you nothing much and it does exactly what you asked.

Point’s has already run by the time you arrive. Ranking happens as mail lands rather than when you go looking. The summary is on the thread before you open it. The ask inside a message has already become a task with a date on it. A request to meet has already had Point running the back and forth, and a real event is on the calendar. Nobody typed a prompt.

Neither is the correct answer in the abstract, and framing this as a client that makes you faster against an assistant that does the work is the wrong frame, which is why an AI email client versus an AI email assistant unpicks it properly. The honest version of the question is about mornings: do you want a capable helper waiting to be asked, or do you want the sorting, the summarizing and the chasing already done when you sit down? An owner who forgets the helper exists gets nothing from the first arrangement, and that is not a small group of people.

The other half of it is what the software may do on its own. In Point that is a setting rather than a decision made in the moment: filing, replying, scheduling and the rest each carry their own level, from suggest-only through review to fully handled, and they all start on review, where Point prepares and you approve. You promote the ones you are ready to hand over, one kind of action at a time, and you can lower them again. Behind it sits an activity log holding every action in order, each explained in a sentence, with undo one click away.

What happens to your mail, and who else sees it?

Spark’s privacy policy is more forthcoming than most, and a firm holding client financial data should read it at source rather than take a competitor’s word for it. Everything below is quoted from that policy, checked 18 August 2026.

It says the data is “mainly processed on your device”, and “may also be processed in Spark systems if necessary for a feature to function”, naming the cases where content is held on its servers, such as an email queued to send later and one shared by link. On the AI side it names the outside services its models may run on: Microsoft’s Azure OpenAI Service, OpenAI, Anthropic’s Claude and Google’s Vertex AI. And it states the thing everybody wants stated: “We do not use your data to train any AI model. AI providers may only retain your inputs for up to thirty (30) days as part of their abuse and misuse monitoring.”

Credit where it is due. Naming the providers and the retention window is more than many vendors will commit to in writing, and a practice can take that policy to a partner meeting and get a decision out of it. That is what a good disclosure is for.

Point’s own answers belong in the same kind of scrutiny, and the questions worth putting to any vendor, Point included, are set out in the confidentiality checks for an AI email client. What is particular to Point is where it draws two lines. Mail from a sender Point does not recognize is held back before anything processes it, so an instruction buried inside a message from a stranger never reaches the AI at all, and releasing it is your call. And the separation Point enforces is between the businesses you operate, so a practice and the thing you run beside it do not see into each other, and a reply leaves under the right one without you picking an account.

Where Spark is stronger

Spark reaches further than anything else on these pages, and the first two items are that reach.

It runs everywhere. Windows, Mac, iPhone, iPad, Android and Apple Watch, per its own site, checked 18 August 2026. Nothing else on these comparison pages covers that much ground.

It connects to anything. IMAP, iCloud, Exchange, Outlook, Yahoo and Google. Point connects to Gmail and to Microsoft 365, so a firm hosted somewhere else is not a Point customer yet.

There is a free way in. Spark’s free tier carries the smart inbox, unlimited accounts and a calendar, which is a real email client at no cost rather than a crippled demo. Point has no free tier and publishes no price during the private beta.

Predictability. Its sorting rule is one you can state in a sentence and check by eye, and nothing is promoted that you did not promote. For a partner who wants to know exactly why a message is where it is, that is a feature and not a limitation.

Maturity. Years of shipping, a large user base, team collaboration built out with shared drafts, comments and delegation, and a privacy policy that commits to specifics. Point is in private beta. Count that gap honestly.

Which one suits a small professional firm?

Ask how much of the deciding you want to keep.

If you know who matters, will happily say so, and want software that does exactly and only what you told it, Spark is built on that principle and executes it well. It also costs nothing to find out, which is not true of anything else here.

If the trouble is that you cannot keep a list of who matters up to date while running a practice, or that the message which hurt you last quarter came from somebody who was never on any list, then a nomination system is being asked to do a job it was not designed for. That is the ground Point is built for.

Choose Spark if your mail lives somewhere neither Google nor Microsoft, if your team is spread across every kind of device, if a free client is what the budget allows, or if you would rather nominate the people who matter than have software judge each message for you.

Choose Point if you want mail weighed on what it asks of you rather than filed by what kind of thing it is, if you want the summary, the task and the arranged meeting waiting rather than available on request, if what the software may do on its own should be a level per kind of action rather than a quota per month, or if you run more than one business and need each kept out of the other’s view.

Not sure a whole client is the right category at all? What an AI email client is settles the definition, what one costs works through the money across the category, and what a first week involves is in switching to an AI email client.

Everything in Spark’s column below was read off its own site on 18 August 2026.

Common questions

Is Point a Spark alternative?

Yes, they are both full email clients and they replace the app you read and send from. What separates them is who does the judging. Spark files by category and then follows the priorities you set; Point weighs each message on what it is asking of you and when, including mail from people you have never dealt with. If the deciding is the part you want to keep, Spark is the better fit.

Does Spark decide what is important, or do I?

Both, in different places. Its Smart Inbox page, checked 18 August 2026, says Spark sorts mail into personal, newsletter and notification categories and puts mail from real people at the top, which is automatic. Which of those real people comes first is yours to set: priority is a mark you apply to senders and threads, and Gatekeeper screens anybody new so you can accept or block them.

Does Spark have a free plan?

Yes, and it is a genuine one. Its pricing page, checked 18 August 2026, offers a free tier with the smart inbox, unlimited email accounts, smart notifications and a calendar. AI is where the paid tiers begin: free users get a limited preview, and the paid plans carry a monthly allowance of AI work. Point publishes no price and has no free tier during the private beta.

Is Spark safe for confidential client mail?

Its policy is unusually specific, which is the right starting point. Checked 18 August 2026, it says data is mainly processed on your device, names the outside AI services that may process the rest, states that your data is not used to train any AI model, and says providers may retain inputs for up to thirty days for abuse monitoring. Whether that is acceptable for your practice is a decision for your partners, and the questions to ask any vendor are in the confidentiality checks.

Which suits an accounting or consulting firm better?

Point, in most cases, and the reason is about who has time. A practice where somebody will maintain the list of important senders and vet each new one gets a precise, predictable result from Spark. A practice where nobody has that afternoon, and where the message that matters keeps arriving from somebody new, is better served by software that weighs the whole inbox without being told who counts. Spark’s platform reach is the honest counterweight, and a firm hosted outside Google and Microsoft should weigh it heavily.

Can Point reply to something without me seeing it first?

At the top of the dial, yes, and having a dial is the point. Each kind of action sits at its own level, from suggest-only through review to fully handled, and out of the box they are all on review, where Point prepares and you approve. You raise the ones you are ready to hand over and lower them again whenever you like, and Point runs the scheduling back and forth on its own. The activity log holds every action in order with undo a click away, though nothing recalls a message once it has gone to somebody else’s server.

The short version

• Spark’s sorting answers what kind of thing a message is. Its own page, checked 18 August 2026, describes personal, newsletter and notification categories with real people put at the top.
• Which real person comes first is yours to declare. Priority is a mark you apply, and Gatekeeper screens each new sender for you to accept or block.
• That is precise and predictable, and it depends on a list built from the past. The message that costs a practice money is often from somebody who was never on it.
• Spark’s AI is invoked and metered as a monthly allowance. Point’s has already run before you sit down, and how far it goes without you is a level you set for each kind of action.
• Spark runs on more devices and connects to more mailboxes than anything else here, and its free tier is a real email client. If your mail lives outside Google and Microsoft, that settles it.

What kind of thing a message is and what it is worth to you are different questions, and the benefits page is the inventory for the second. Accountants, law practices and everyone else get their own version of the answer.