What's New

Keep up with what’s changed in ProcessPlan. Entries you haven’t seen since your last visit are marked New.

My Tasks keeps you on your default viewAugust 18, 2026

My Tasks leaves out tasks scheduled to start on a future date, so the list stays the work you can pick up now. Clicking My Tasks in the left-hand navigation used to drop that default the second time you clicked it, quietly padding the list with work you cannot start yet. Now it holds, and so does anything you filtered for yourself: you come back to the list as you left it. Switching a tile such as Due Today back off, or hiding the filter row, returns you to the default view. See Viewing Your Tasks.

Findings close themselves when the situation improvesAugust 17, 2026

Every analyst now re-checks the findings it has already raised on each pass, and closes one the moment the numbers behind it come back inside normal range, recording what changed. Findings that are still live refresh with the latest measurement, so you can see whether one is getting better or worse since you first saw it.

When an analyst cannot take a measurement at all, because the process it watches has no process assigned to it, a field it measured has been removed, or there is too little recent activity to judge, it now says exactly that on the finding instead of going quiet. It never closes a finding it could not measure, so a closed finding always means something genuinely improved. See Focus Areas.

Your AI Strategist briefing, on whatever schedule suits youAugust 15, 2026

The briefing used to arrive on one fixed day. Now you say when you want it and the Strategist sets it up in a strategy session: “email me the briefing every morning”, “just weekdays”, “every Friday”, “the first Monday of the month”. It reads the schedule back with the next few dates it will arrive, so you can correct it on the spot, and it tells you plainly if you have asked for something it cannot arrange rather than quietly settling for something close. The common choices are in your profile settings too, under AI Strategist Briefing, once Send AI Strategist Emails is switched on.

Choosing a daily briefing does not mean a daily email. It stays quiet whenever nothing has changed since the last one you read.

You can also ask for a briefing about your own corner of the business: “I want to keep an eye on how onboarding is going”. The Strategist watches what you described and tells you about that, limited to the work you already have access to.

Nothing signs you up on your behalf, and nothing changed for anyone already subscribed: existing subscriptions carry on arriving exactly as they did. Send AI Strategist Emails in your profile turns off every email from the Strategist at once, and everything is still there in the product whenever you want it. See The AI Strategist.

Public forms can say the person's own details back to themAugust 15, 2026

Instructions and Success Message or Redirect URL on a public form now understand Field Tokens. Insert them the same way you would anywhere else and the person filling in the form sees real values, not token text.

On a form that updates a process already running, the instructions can open with what you already have on file: “We have your delivery address as 41 High Street. Change anything below that is out of date.” In the success message, tokens are filled in from the process that was just started or updated, so you can confirm their details back to them or give them a reference to quote later. A token with nothing to fill in is left blank.

Text Blocks are now Account Values, and they can hold more than textAugust 15, 2026

Reusable Text Blocks have a new name and a new job. Account Values still do everything text blocks did, and every one you already have keeps working exactly as before, under its existing name. What is new is that a value no longer has to be a formatted block of text.

When you create or edit an Account Value, you now choose a Value type: Rich Text for signatures and disclaimers, or Plain Text, Email Address, URL, or Number for the single pieces of information you would otherwise retype across a dozen templates. Picking a plain type gives you a simple one line box instead of the full editor, and the value arrives clean wherever it is used, including places that cannot show formatting, such as a text message.

In the field token list, the tokens are now named Account Value: followed by the value’s name. Anything you have already saved keeps working untouched.

Calculations gain powers, decimal remainders, and much bigger numbersAugust 13, 2026

The =CALC formula has a rebuilt calculator behind it. Everyday arithmetic returns exactly the same answers as before, including money, where =CALC(0.1 + 0.2) is 0.3 and =CALC(1.10 * 3) is 3.30. What is new:

• ^ raises a number to a power. =CALC(2 ^ 3) is 8, and a fractional power takes a root, so =CALC(9 ^ 0.5) is 3.
• % gives you the remainder left after dividing. It was never documented before, and it now handles decimals, so =CALC(10.5 % 3) is 1.5. It also sits alongside * and / in the usual order of operations.
• Large numbers no longer come back blank. Multiplying values into the billions and beyond now works.

Two results changed. Dividing by zero returns nothing instead of the word Infinity. And a chained comparison like =CALC(1 < 2 < 3) now returns nothing rather than an answer that looked right but was not, so write =CALC((1 < 2) AND (2 < 3)) instead.

The full operator list, including the comparison operators, is on the =CALC page.

Four analysts that were staying silent now speak upAugust 12, 2026

Four focus areas compare a run against the normal pattern for its process, and that comparison was failing, so they finished every check with nothing to say. If you turned one of these on and it never spoke up, this is why, and there is nothing for you to change: the next check runs properly on its own.

• Post-Mortem Debriefs, which studies the runs that ended badly and asks the people involved what actually happened.
• Institutional Memory, which recognizes a customer, vendor, or site at kickoff and tells you how work for them has gone before.
• Supplier Risk and Customer Churn Risk, which weigh what is happening to a supplier or customer in the world against how they behave inside your processes.

Pre-Mortem Guard was speaking up already, but its warnings were missing one comparison: how the process’s past large runs actually ended against the rest. That figure is back in its briefings now.

Supplier Risk and Customer Churn Risk also have full write-ups in Meet the analysts for the first time.

Min and Max fields can read dates, and aggregate columns filter properlyAugust 11, 2026

Process Table Min and Max fields now come in two kinds: (Number) and (Date). The date kind shows the earliest or latest date across the runs that match, so a process can carry the next renewal date on a contract, the last time an order shipped, or the first appointment still open. Pick the kind you want and the list of fields to read from narrows to the ones it can actually read, so a mismatch is no longer something you find out about when you save.

The date is shown in your account’s date format, and when nothing matches the field stays empty rather than showing a zero. Existing Min and Max fields keep working as the (Number) kind.

Filtering these columns in a table view also works properly now. A date column offers the same date-range picker as any other date column, with presets like This Month and Last 7 Days, and Sum, Average, Count, Median and Distinct Count columns offer a number range. Previously any of these columns gave you a plain text box that quietly matched nothing.

Diagram search now takes you to the match, and a new overview mapAugust 10, 2026

Searching a process diagram used to outline the matching steps in red and leave you to find them. On a wide process the matches were almost always off screen, so typing in the search box looked like it had done nothing at all. Now the diagram travels to the first match and zooms in far enough to read it, with a 2 / 7 style counter in the top-left corner of the canvas and ‹ › buttons to step through the rest in flow order.

The canvas also has a new overview map in its bottom-right corner: the whole diagram in miniature, with a frame showing the part you are looking at. Drag inside it to move somewhere else in the process, or scroll on it to zoom. It shows up only while part of the diagram is off screen, so on a process that fits on one screen there is nothing extra in the corner. Both the map and the search jump work on template diagrams and on the diagrams of running processes.

Loop-back connectors stay close to the steps they return pastAugust 10, 2026

A connector that loops back to an earlier step now drops only as far as it needs to clear the steps it actually passes, instead of travelling below the whole diagram. On a wide process, a short return between two neighbouring steps used to dive down past the deepest column anywhere on the canvas and climb back up, which made the diagram much taller than it needed to be and turned every return into a long detour.

A return still never routes through the steps it passes, and where several of them cover the same stretch of the diagram they still stack into their own rows with the shorter returns nearest the steps. The amber routing handle still gives a line more clearance below those steps, within a limit: it won’t move up into the steps, and it won’t drift far below them either. Existing diagrams pick up the tighter routing as soon as you open them.

A JSON answer from an AI Prompt action is stored on its own againAugust 10, 2026

When an AI Prompt action answers with JSON, only the JSON is stored in the field or matched against a task response. That trimming was being skipped whenever the AI added a closing remark that happened to contain square brackets, such as a note that a value in the document did not match one of the choices listed in the prompt. The field ended up holding the code block and the remark along with the answer.

The AI Prompt page now also spells out a shortcut that was never written down: ask for JSON with exactly one property and the value of that property is stored, so {"decision":"Approved"} puts Approved in the field. That is the cleanest way to let the AI pick a task response and drive a decision tree.

=JSONEXTRACT reads the same way, and its page has been corrected: when text holds more than one JSON structure, the function returns the last one, not the first.

Copy or move only some of a field's attachmentsAugust 9, 2026

The Field: Append Value action now works on file attachment fields, adding files to a field without disturbing the ones already there. Previously only Field: Replace Value could copy an attachment, and it always replaced everything in the destination.

Both actions now also offer an Only These Files box when the field they are updating holds files. Leave it empty to take every file, or enter a name with * or ? wildcards, such as *.pdf, to take only the PDFs and leave the images where they are. Pair a copy with a Field: Remove Value on the source field to move files rather than copy them, all covered in copying attachments between fields.

ConnectWise boards now narrow the lists below themAugust 8, 2026

On a process that carries ConnectWise fields, choosing the ConnectWise Board now narrows the ticket status, type, sub type, and item lists to that board straight away, the same way choosing the company has always narrowed the contact, opportunity, ticket, and agreement lists. Choosing a sub type narrows the item list again to the items that sub type allows.

Previously those board-driven lists only caught up with the board once the form was reloaded, so people could be offered statuses and items belonging to a different board.

The ConnectWise integration, documented end to endAugust 8, 2026

There is now a single page covering everything the ConnectWise Manage integration does: the credentials you need, the process field types that search your live ConnectWise companies, contacts, tickets, agreements, boards, and members, how those lists narrow each other, and the actions that create tickets and log time entries for you. It ends with a worked example and a list of what to check when a list comes back empty.

Send an analyst's findings to whoever owns the recordAugust 7, 2026

Every analyst that watches a process now offers an Also notify the people named in this field setting, listing that process’s people fields. Point it at the field holding the account manager, the category owner, or whoever else is named on the record, and a finding about that record reaches them as well as everyone on the watcher list. Each finding goes to the person who owns it instead of to one fixed list.

This was previously only available by asking the Strategist for it. It is now a setting like any other: it appears in the analyst’s configuration, you can still just ask for it in a strategy session, and the analyst can raise it with you itself when it would help. Leave it blank and only the watchers are told.

Row Repeater tokens are now easier to pick outAugust 7, 2026

A Process Table Lookup field that can match several records offers a (Row Repeater) version of each field it pulls through, so an email can print one table row per matched record. Where the field it pulls through is an email address field, that repeater token used to be listed under the same name as the plain one, leaving two identical-looking entries in the token list with no way to tell which was which. It is now labelled (Row Repeater) like every other one.

The token list is also the only safe way to insert a token, so this is worth knowing about: see Field Tokens for how row repeaters work.

Loop-back connectors now always run below the diagramAugust 5, 2026

A connector that loops back to an earlier step now always travels back underneath the whole diagram rather than through the middle of it. Previously a return line only dropped far enough to clear the steps it passed directly over, which could leave it threading between two rows of steps where it read as part of the forward flow. Where several loop-backs run at once they stack into their own rows below the diagram, shorter returns nearest the steps, so nested returns nest instead of crossing.

The amber routing handle still works: drag it down to give a return line more clearance below the diagram. It no longer moves up into the steps. Existing diagrams pick this up as soon as you open them, with no need to re-save or re-format.

Ask the AI Strategist to hand someone's tasks overAugust 5, 2026

When somebody leaves, goes on leave, or hands a role over, you can now ask the AI Strategist to move their open tasks in a strategy session: “Priya has left, move her open work to Dana.” Before it does, it shows you the list - which tasks would move, who holds them now, and any that cannot be moved along with the reason, such as a task that has already been answered or a step whose design does not allow reassignment. Nothing changes until you say yes, and the new assignee is notified exactly as they would be for any other assignment.

Name particular tasks, or ask it to take everything one person still holds and optionally narrow that to a single process. Because moving the tasks that exist today does not change where tomorrow’s go, the Strategist also offers to change the step’s assignment, which is usually what a permanent handover needs. You need permission to change task assignments, the same as reassigning a task yourself.

Departure Shield now offers the handover, not just the list. When a user is removed from your account and their record names someone who can pick up their work - the backup they designated, or the person they reported to - the Departure Shield card arrives with the handover ready to approve: one approval moves their whole open queue to that person. It re-reads the queue at the moment you approve, so anything already picked up by hand is left alone, and it says plainly if any task could not be moved and why. Approving it needs permission to change task assignments, the same as moving those tasks by hand. When the record names nobody, it does not guess - the card stays the checklist it was.

Approving an action now needs the permission that action needs. Where the Strategist recommends something and carries it out for you, approving it takes the same permission you would need to do it yourself: moving a step to someone else needs permission to change task assignments, and cancelling a run needs permission to cancel processes. Previously permission to guide the AI Strategist was enough on its own for those two. If you do not have it, the recommendation stays put for someone who does, and the Strategist says so.

Copy a file attachment into another processAugust 4, 2026

The Field: Replace Value automated action now works on file attachment fields. Point it at the attachment field you are copying from using that field’s (Value) token, and the file lands in the field you are updating, including when that field lives in a linked or looked-up process instance. The file is shared rather than duplicated, so it is one document appearing in both places.

Two related improvements. If the value you are replacing with turns out to be empty, the action now fails and leaves the existing files alone instead of quietly emptying the field, and Field: Clear Value remains the way to deliberately remove attachments. And this action now reports failure honestly: when it cannot find or update the field it says so in the audit trail and runs any actions you have triggered on failure, where before it always claimed success.

Set up anything about an analyst just by askingAugust 3, 2026

Everything about one of the AI Strategist’s analysts can now be set up and changed in a strategy session, so there is nothing about one you have to go and configure on a settings screen. Ask how an analyst is currently set up and it tells you the whole picture, including the standing guidance you have given it over the months. Ask for a change and it makes it: rename it, retarget it, change how often it checks or how conservative it is, change who hears about its findings, or have it also notify whoever is named on the record itself.

Two of those are worth calling out. When you narrow which records an analyst follows (“only the customers whose status is active”), the Strategist works out the conditions, runs them before saving anything, and shows you what they actually selected: how many records matched out of the total, and the most recent ones by name. You confirm the set is right before it is saved, which catches the filter that quietly matches nothing long before it costs you weeks of wrong findings.

And you can now tell an analyst what should happen when it finds something: which of your processes should run, and which of that process’s fields should receive the finding’s headline, summary, detail, the process it concerns, the people to notify, and the record it is about. Then choose whether it tells you only, or brings the action for your approval. The Strategist will not put itself in charge: ask it to start acting without approval and it declines until it has a track record of approved actions with measured improvements behind them, telling you where that record stands. An administrator can still grant it directly in the analyst’s settings, and either route is recorded in your decision journal.

Two more things you can now ask for. Quiet an analyst until a date (“stop flagging late shipments until January 5th”) and it goes silent, then picks itself back up on that date, so there is no paused analyst waiting for someone to remember it. And withdraw guidance that no longer applies: tell the Strategist that something you told it before has changed and it retires that instruction instead of stacking a contradiction on top, says exactly what stopped applying, and keeps the history in your decision journal.

See Focus Areas.

Promised dates are measured at the step you actually promisedAugust 3, 2026

The Commitment Reliability analyst judges how often you hit the dates you give customers. A promised date is often about one step, not about the whole run finishing: you promise the delivery, then the run carries on with invoicing and closeout. Measured against the end of the run, a delivery you made early can look weeks late.

Now you can tell the analyst which step has to be done by the promised date, and it judges the promise the first time that step is completed. Everything after it stops counting against you. You do not have to know to set this up: when the numbers show a run consistently continuing well past its promised date while one step lands right on it, the AI Strategist says what it suspects and asks you which step it is. Name the step in your reply and it changes how it measures; tell it the promise really is about the whole run and it leaves things as they are and stops asking.

It also now names the promises that are already past their date with nothing delivered yet, so the ones you can still rescue arrive as something to act on rather than as history.

The AI Strategist asks who to go toAugust 2, 2026

To act on something, the AI Strategist often needs to reach the right person, and a list of names does not tell it who handles what. Now, when it needs to raise something and cannot see who owns it, it asks you right then and says why: “I need to ask about product pricing and I can’t see who owns that. Who should I ask?” Tell it, and it notes down what that person does and stops asking. It also tells you how it worked out who to approach, so you can correct it when it guesses wrong.

You do not have to wait to be asked. Administrators can set Functional area and Seniority level on anyone’s record under Administration > Users, and both are included in the user list CSV export so you can see at a glance who is still missing them.

The AI Strategist is now included with every subscriptionAugust 1, 2026

The AI Strategist is part of ProcessPlan for every customer. There is no longer an add-on to turn on and no separate charge for it, and the toggle has been removed from the Plan & Payment page. If your account was not already using it, the Strategist starts work on its own: it looks over how your account runs, sets up the focus areas that fit your data, and begins filing its AI Daily Briefing. Nothing to configure or approve first.

Cross-Company Peer Insights is included too. It stays opt-in, because it means sharing anonymized benchmarks with other companies, but opting in no longer adds anything to your bill. You will find its toggle in the same place.

The Strategist works for any account with an active subscription or a trial in progress. If a subscription lapses, its analysts stop, your focus areas and briefing pages say so, and everything picks straight back up once billing is current.

The Strategist hires its own analysts, without asking firstAugust 1, 2026

No focus area waits for your approval to start. When the Strategist sees a pattern one of its analysts is built for, it puts that analyst to work and tells you it did, rather than queuing up a request for you to approve. The standing “ready to staff itself” offer is gone, because it was asking permission for something already happening.

Every analyst in the catalog now has a rule for when it is warranted, so the roster fills itself in as your data grows: the per-entity watchers once your processes reference the same customers repeatedly, the outcome lenses once your responses carry ratings, the statistical ones once there are enough completed runs for the maths to be honest. Each starts observe-first, each is recorded in the decision journal, and any of them can be paused or retired with a word in a strategy session.

Approval still belongs where it matters: on what an analyst proposes to do. A change to a process, work launched on your behalf, or a step automated still waits for you every time. Two exceptions keep asking first, both because they reach beyond your own data: the watchers of outside companies ask which records to follow before researching them on the open web, and handing a process over to the AI Process Owner stays your call, since that one runs the queue and contacts your people. See Focus Areas.

Point a file search straight at the folders you care aboutJuly 31, 2026

Every AI Employee action that searches for files now accepts a pattern in the folder path as well as the file name. Wrap the part you want matched in angle brackets and leave the rest as ordinary text: C:\Sales\<20\d\d>\Invoices searches the Invoices folder inside every four digit year folder, and report_<\d{4}>.pdf matches file names such as report_2026.pdf.

Searches over big folder trees get much faster, because the AI Employee only opens the folders that match instead of reading through everything underneath the starting folder. Angle brackets are not allowed in Windows file and folder names, so an ordinary path is never mistaken for a pattern and there is nothing to switch on. This replaces the File Names / Regex (includes subfolders) switch; if you were using it, put the pattern in the folder path or file name instead, and turn on Search All Subfolders where you need it. See Matching Folders and File Names.

One related change: Files and Folders as JSON Array now returns at most 100 entries, in name order, and says in the log when the list was cut short. That result is meant to be read by an AI prompt or a script, and a longer list is rarely useful to either. If one of your jobs was relying on a longer listing, narrow it with a pattern in the folder path so the list you get back is complete.

Ask the AI Strategist to put an AI agent on a step and it sets it up for youJuly 31, 2026

Ask the AI Strategist to have AI handle part of a process - drafting the reply, summarizing the document that came in, researching a request, classifying it - and it now sets that AI Agent action up inside the conversation. Previously it would explain where to configure one and leave the setup to you.

It shows you exactly what it will put in place first: which step, what the agent will do, and which field its answer gets written into so you can read and edit it. Say yes and the automation is live straight away. It runs the next time that step’s task is assigned.

The agent does the work and then completes the step, choosing the response that fits what it found - the success route when the work came out clean, the failure or exception route when it did not. That choice is how your team sees both that the work is finished and how it turned out, and the run carries on down whichever route that response leads to, exactly as your diagram lays it out. So tell the Strategist which response should be chosen in which circumstance and it writes that into the agent’s instructions. If you want a person to check the agent’s work before the run goes further, that is simply a review step after the agent’s step, and the Strategist can add it for you in the same conversation.

If a step already has an AI agent on it, the Strategist tells you what is running there rather than stacking a second one on top. Setting this up requires edit access to the process.

The AI Strategist works out what an automation will cost before you build itJuly 30, 2026

Describe a job you would like an AI Employee to take over and the AI Strategist now tells you what it would cost to run each month. It works before anything has been built and before the job has ever run, so you no longer have to convert your team’s hours into a bill in your head to find out whether automating something is worth it.

It breaks the job into its real steps and works out which parts need an AI Employee working a screen and which can be handled by AI reading documents and data instead, which is usually the cheaper way to do it. You get a range rather than a single figure, along with the assumptions it had to make, so you can correct anything that is wrong and get a sharper number straight away. If the work fits within the hours already available on a workstation you are paying for, it says so instead of quoting you for another one.

Ask it directly, as in “what would it cost to check every invoice against its purchase order?”, or let it raise the number itself when a job worth automating comes up in conversation. For work that needs a screen recording, it gives you a rough setup range as well; the exact one-time fine-tuning price is still quoted from your own recording, before you commit to anything.

AI Employees check documents exactly instead of going from memoryJuly 28, 2026

Ask an AI - File Related Prompt whether particular values appear in a PDF, text, or CSV file and each value is now checked exactly against the text in the document. Long reports full of similar looking codes and reference numbers were the weak spot before: reading a hundred page document and recalling whether one account number out of a thousand was present is exactly the kind of question that could come back differently on two runs of the same job. Now you can hand over a whole list and ask which entries are missing, or which is the first one missing, and rely on the answer.

The AI Employee also takes a little more time to reason through file questions, which shows up as better answers on anything that takes several steps: comparing a list against a document, cross referencing two sections, or working out totals from a table. If a PDF turns out to be only a scanned picture of a page, with no text to check, it now says so instead of guessing.

AI Prompt actions handle attached spreadsheets and files betterJuly 28, 2026

Dates in an attached Excel spreadsheet now reach the AI as real dates. A spreadsheet that used anything other than Excel’s default date formats could previously hand the AI a raw number like 45678 to make sense of, which is why the occasional answer got a date wrong. Times, and cells holding an elapsed duration such as a timesheet total, come through accurately now too.

Separately, an AI Prompt action now treats the contents of an attached file strictly as material to work on, never as instructions to obey. Text that tells the AI what to do is not followed if it sits inside an attached file: in a spreadsheet cell, in a document, or as text in an image. This matters because attachments often reach a process from outside your company, through a public form, an emailed reply, or a file a supplier sent you. The subject matter of a file is still read normally, so an attached policy that says “approve when the total is under $5,000” is information the AI will use when it answers your question.

The screen recorder now reminds you to narrate instead of ending your recordingJuly 28, 2026

The screen recorder used to end a recording if it stopped hearing you, which meant a long stretch of quiet concentration could cost you the whole take. It no longer does that. Once the recorder has heard your voice, silence never stops the recording - if you go quiet for 30 seconds you get a Narration Needed reminder in the corner of your screen, repeating every 30 seconds until you start talking again. Your recording keeps running the whole time, and nothing appears in the captured video.

The recorder also speaks up about a problem it used to keep to itself. If it can hear you but your narration is too quiet to be processed, you now get a Microphone Too Quiet reminder on the same 30-second rhythm. Previously a too-quiet recording ran happily to the end and only disappointed you afterwards.

The one check that still stops a recording is at the very beginning: if the recorder hears nothing at all during the first 30 seconds, it stops and tells you, because that nearly always means the microphone is muted or the wrong one is selected - and there is no point recording a process nobody can hear. Pausing no longer counts against you either: step away with Ctrl+Shift+P and you are not reminded, and the paused time is not held against the opening 30 seconds. Restarting a recording now keeps the microphone you picked instead of quietly reverting to the system default.

The AI Strategist checks its work before changing a process, and can now reshape the fields you already haveJuly 27, 2026

Ask the Strategist to change a process and it now tests the change against your live process before anything is saved, telling you if the result would not hold together - if removing a step would leave later steps with nothing leading to them, you hear about it while the process is still untouched. It can also do far more to the data a process captures: alongside adding fields it can rename them, make them required or optional, reorder them, add dropdown choices, and point a process table lookup at a different table or narrow which records it offers. It shows you exactly what it intends to change and waits for your yes. A few things it will always decline, and say why: changing an existing field’s type, deleting fields, and repointing a Sum or Count field.

The Quick Diagram Builder has been retired. If you have your steps written down as a list, paste them into a conversation with the Strategist and ask it to build the process - it proposes the steps and outcomes, and builds once you agree.

Diagram connector lines no longer hide behind each otherJuly 27, 2026

In the diagram editor, connector lines now keep out of each other’s way. Where two lines used to run along the same path, one drawn directly over the other so what looked like a single connector was really two, they are now offset so both stay visible. Lines still cross where the flow calls for it. Where several connectors leave or arrive at the same step they fan out into their own lanes instead of stacking up as one thick line, and if you drag a loop-back line’s routing handle right onto another connector, it settles just above or below that line rather than disappearing behind it. Your layout is untouched: steps stay exactly where you put them, and only the lines between them shift enough to stay distinct.

The AI Strategist can now fix a lookup field that offers the wrong recordsJuly 27, 2026

A process table lookup pointed at the wrong table shows up as a field that is mysteriously always empty. Tell your Strategist about it in a strategy session: it now checks which table the field reads, and once you confirm the table it should read instead, it makes the change itself. It can also set or remove the filter that limits which records the field offers - including a filter that follows a value already entered on the same form, so a Project field can offer only the projects belonging to the company chosen a moment earlier. Values already saved on runs in progress are left exactly as they are.

Ask the AI Strategist how your numbers are distributed, and let it watch for unpredictable timelinesJuly 26, 2026

Two additions for anyone who plans against their process data. In a strategy session you can now ask how a number is actually distributed - “what does our order-value spread look like?”, “how long does fulfillment take, best case and worst case?” - and the Strategist computes the typical value, the spread, the slowest 1 in 20, and the shape from every matching run. If you build simulation or forecasting models, it will even tell you which probability distribution fits a variable best, with the parameters to plug in. And a new Cycle-Time Predictability analyst watches for the process that runs steadily but finishes on a timeline nobody can promise against - a typical run in days, the slowest in weeks - and points at what the slow runs share so the spread can be tightened for good.

Click On Text joins the action list, and AI-built automations wait for your reviewJuly 25, 2026

The simplest computer vision click, Click On Visible Text, previously appeared only in jobs the AI built from a screen recording. You can now add one by hand: choose Computer Vision - Click On Text from the Action Type list. Also, when the AI builds an automation from a screen recording, the automation now arrives switched off so you can review the generated actions first; it will not trigger on its own until you turn it on.

The AI Strategist now proves improvements with real statistics and guards the gainsJuly 25, 2026

The Strategist’s judgment got a serious upgrade in rigor. A new Process Stability analyst learns the range each process normally runs in and speaks up only when the numbers genuinely leave it - ordinary ups and downs stay quiet, and favorable shifts are flagged too so what caused them can become standard practice. When you approve an improvement, the Strategist now verifies it statistically: it measures before and after, only calls a gain real when it clearly beats the process’s normal variation, keeps watching afterward, and raises it if the gain ever starts slipping. And you can now ask the Strategist how reliably any process is running - it computes the answer from your real runs and says honestly when there is not enough history yet.

Your AI Daily Briefing now organizes around your strategic initiativesJuly 25, 2026

The AI Strategist now connects every finding its analysts make to the strategic initiative it advances or threatens, and your AI Daily Briefing leads with them: findings that bear on a stated priority come first and name the initiative they serve, decisions waiting on you are labeled with the initiative they affect, and when nothing is moving on one of your initiatives the Strategist says so plainly. Findings unrelated to any initiative still surface on their own merits - initiatives decide what leads, not what gets watched.

Hand a process to the AI Strategist to run end to endJuly 22, 2026

With the new Process Ownership capability you can hand one process to the AI Strategist and it runs that process’s queue for you: it watches every open run, checks in personally with whoever an overdue or unusually slow step is waiting on, follows up if nothing moves, and escalates a genuinely stuck run to you with the right decision teed up. Before it escalates, it asks the assignee directly what the step is waiting on (they answer through an emailed link that opens a conversation with the Strategist about that exact item), and their answer picks the recommendation: reassign the step to the person who has done it before with the lightest queue, extend the deadline when the blocker is external or time-bound, or cancel the run when the work is no longer needed - each carried out by the Strategist itself on your one approval. When an owned process’s typical run time drifts well past the baseline recorded at handover, it opens an improvement campaign: concrete changes pulled one at a time (a deadline policy on the bottleneck step, a rebalance of where that step’s work lands), each made on your approval and measured on its own before-and-after, until the number is back or it says plainly that it has exhausted its levers. And for solutions that take more than one piece, you can now design a complete solution with the Strategist in conversation - a new process, changes to existing ones, and automations connecting them - and it files the whole plan as one proposal: a single approval, given right in the conversation or from your Decision Queue, builds every piece in order, and if a step cannot be built it stops and tells you exactly what was completed. Approved moves build a track record that can earn it a standing offer to route stuck steps on its own, with hard guardrails: at most one unassisted move per step, every move journaled, and runs you are already deciding on stay yours. Results are measured, not asserted: it records how the process performed before the handover and its owner’s report in the AI Daily Briefing compares runs on its watch against that baseline, saying “too early to measure” until the data is there. Ownership is always your explicit decision (approve its offer, ask for it in chat, or add the AI Process Owner focus area by hand), it never changes how the process works without your approval, and you can take the process back anytime.

File actions can match names with a regular expression, including subfoldersJuly 20, 2026

Every AI Employee action that searches for files by name - moving, copying, finding, opening, deleting, listing files as JSON, starting an instance for each file row, and uploading files into a process field - has a new File Names / Regex (includes subfolders) switch. Leave it off to keep matching file names with * and ? wildcards and semicolon-separated patterns. Turn it on to write one regular expression that is matched against each file’s path relative to the folder, so a single pattern like .*\.pdf$ can find files in the folder and every subfolder beneath it.

This switch was replaced on 31 July 2026 by patterns you write directly in the folder path and file name. See Matching Folders and File Names for how it works now.

Ask the AI to change who a step is assigned toJuly 19, 2026

In any AI conversation, including a strategy session, you can now ask to reassign a step in one of your processes: to a specific person, to a team as one shared task, to every member of a team individually (each person gets their own copy), to whoever on the team has the fewest pending tasks, or back to the process initiator or the previous step’s owner. The AI confirms the step and the new assignee with you before making the change, and the change applies to tasks created from then on; tasks already in flight keep their current assignee.

Automation upgrades: dynamic table searches and sharper conditionsJuly 19, 2026

The Process: Find In Process Table action now accepts field tokens in its Process Table Query, so a search can be driven by live values from the current process instead of fixed text. Automation conditions got sharper too: conditions that compare a User List field to a specific person, and conditions on Process Table Lookup fields, now match exactly and predictably in both their “equals” and “contains” forms.

Run the AI Strategist's whole team from the conversationJuly 18, 2026

You no longer need a configuration screen to direct the AI Strategist. In any strategy session, ask it to put a new focus area to work and it resolves the details from your data, plays the exact plan back, and sets the analyst up once you agree - observe-first, always. Pausing, retargeting, changing settings, and asking for more or less sensitivity all work the same way. Your direction sticks: standing facts and rules you state in conversation are recorded (with your confirmation) and honored by every future analysis. When a focus area fits your data but needs one answer only you can give, the Strategist asks the one question instead of staying silent. It learns from dismissals too - never re-raising equivalents of what you dismissed, and checking in once when several dismissals share a pattern. And once you’ve directed a few setups yourself, it offers to staff the clear-cut analysts on its own: announced after each one, reversible with a word in chat, and never without your standing approval.

New specialists join the AI Strategist's teamJuly 17, 2026

The AI Strategist’s roster of focus areas just grew. New analysts plan coverage before someone goes on vacation, act the moment an employee leaves, notice when today is running far off your normal rhythm, remember how past work for a customer went and say so at kickoff, catch the moments that go unusually well while a thank-you still lands, prepare your side of a negotiation from your own records, watch the outside world for RFPs and grants that fit your company, quietly notice when someone’s workload has been climbing for weeks, and more. The Strategist sets up the ones that fit your data itself; browse the catalog to add the rest. See Focus Areas to meet them.

Tell the AI Strategist a process is retiredJuly 16, 2026

When a process is no longer part of how you work, just say so in any strategy session: the Strategist marks it retired, with your confirmation, and stops watching it, so you won’t see suggestions or proposals about it anymore. The same works right while reviewing one of its recommendations: tell it the process is deprecated and it declines the recommendation and offers to retire the process on the spot. Nothing is deleted, and it can bring the process back if you change your mind. You can also ask it to pause or resume any of its focus areas from the same conversation.

Document a process by recording your screenJuly 14, 2026

The new ProcessPlan Recorder for Windows captures your screen and your spoken narration while you perform a process the way you normally would. Stop the recording and it uploads automatically, and ProcessPlan turns it into a documented process, either as a new process or merged into one you already have. See the Screen Recorder guide to get started.

Connect AI assistants to ProcessPlanJuly 14, 2026

ProcessPlan now speaks MCP (Model Context Protocol), the open standard AI assistants use to connect to business tools. Point any MCP-capable assistant at your account with an API token and it can browse your process templates, check on running processes, review tasks and field values, and complete tasks for you, all with your user’s permissions. See Connecting AI Assistants (MCP) for setup instructions.

The AI Strategist builds what you agree onJuly 13, 2026

A strategy session no longer ends with a to-do list. When you and the AI Strategist settle on a change, it can carry it out right from the conversation: create a new process or data table, add steps or outcomes to an existing process, add new data fields to an existing process (new), and launch or schedule process runs. It always shows you the exact design and asks for your confirmation before it builds anything. The Strategist’s setup interview for new accounts builds the same way: confirm a process design in the conversation and it is created on the spot, ready to run, with no approval step in between.

A diagram editor that keeps your layoutJuly 13, 2026

The process diagram editor has been rebuilt. Add steps with the amber + button, draw connections between steps, reroute loop-back connectors around other nodes, and annotate the canvas with sticky notes. See The Diagram Editor for the full tour. Best of all, your diagram now keeps its shape: the layout you design is saved exactly as you arranged it and is never reformatted automatically, and the same shape carries through to the diagrams of running processes and heat maps.

Redesigned date and number filtersJuly 12, 2026

Date columns in tables now filter with a date range selector: pick a preset like Last 7 days, This Month, or Last Quarter, or choose a custom From/To range. Presets stay relative, so a bookmarked or shared view always shows the current period. Number columns get a matching minimum/maximum selector. See Filtering Columns for the details, and Filtering Lists for the new date range syntax in the API.

New documentation siteJuly 12, 2026

Welcome to the new ProcessPlan documentation site! It covers the core of the product (process templates, running processes, tasks, process diagrams, and the AI Strategist) with full-text search (press / or click the search bar). We’ll post new features and release notes here as they ship.