MAX.AI Book a call

BUSINESS PROCESS ENGINEERS

The output of a department. The headcount of one.

We map how a business actually runs, department by department, then rebuild the processes that should not need a person doing them.

Book a call Walk the map

14
systems built
6
live in production

Garden maintenance · Property acquisition · Serviced accommodation · Construction recruitment · Technology recruitment · Hospitality finance recruitment · Planning consultancy · Lettings management

Problem

Every business over five people has the same shape of problem

Not the same problem. The same shape. Work arrives, a person moves it to the next person, and somewhere in that chain one human being has quietly become the integration layer. Everything you own notifies you. None of it does anything.

Audit

We map how it actually runs, not how you describe it

One session, screen shared, walking the real thing: the tabs, the group chat, the spreadsheet only one person understands. Owners consistently discover something in that hour they did not know was true about their own business. That is the point of it.

Systemise

Order first. Automating a mess just makes a faster mess.

Before anything is built, the sequence gets fixed: what has to happen, in what order, and where a human genuinely has to decide. Most of the value shows up here, and it shows up before you have paid for a single line of code.

Automate

Only the part that truly does not need a person

The chasing, the copying, the re-typing, the remembering. Not the judgement. Every system we build stops and waits where a person should decide, and nothing goes out in your name that you have not seen.

Solve

The thing that used to fail stops failing

That is the whole test. Not hours saved on a slide. Something that used to break, or leak, or get forgotten, measurably does not any more. If we cannot point at that, we have not finished.

Own

It runs on your infrastructure, in your name

Your domain, your accounts, your data. No seat fee, no platform holding your operation hostage, no dependency on us to keep it running. You own the thing we built, which is the only arrangement worth having.

WHAT THIS LOOKS LIKE

Fourteen businesses, told from the inside

Each of these begins with a problem you may recognise. Read it as your own. The last panel tells you whose it actually was.

Case study: Property acquisition and investment

A property firm whose best leads arrive cold / The problem

The best enquiry you get is what is it worth

Someone sitting on a block worth seven figures wants a number. It is the single most valuable question anybody asks you, and answering it properly costs a real slice of somebody's afternoon: comparables, rents, yields, a defensible range, a document that does not embarrass you. So you answer the ones you have time for. The rest wait, cool off, and go and ask somebody else instead.

A property firm whose best leads arrive cold / What we found

Free valuations are only free if defensible

Mapping it surfaced the thing that makes an instant valuation dangerous rather than clever. You carry professional indemnity cover, and a number that reads like a formal valuation walks straight into it. Thin comparable data is worse than none, because a machine will still produce a confident figure from nothing. The hard part was never the maths. It was the refusal: knowing when to give no number at all.

A property firm whose best leads arrive cold / What we built

A number you would stand behind, in seconds

A vendor fills in a short form and, in about half a minute, has a branded appraisal and a full investment memorandum in the inbox: cautious, core and optimistic, on the correct basis for the asset, with the evidence and the assumptions written out underneath. Out of your area, or too complex to price fairly, and it declines politely and asks for a conversation instead. Every enquiry arrives on your board already scored, tagged and consented.

A property firm whose best leads arrive cold / What changed

Valuations stopped being a diary problem

It runs at any hour with nobody at a desk. It went into real use fast enough that the first round of change requests were all about live behaviour rather than ideas: copy every report to the team automatically, split the form into two steps so fewer people abandon it halfway, move the headline figure into the attached memorandum so a vendor opens the document instead of reacting to one line. No conversion numbers have been shared, so none are claimed.

~30 secondsFrom a vendor submitting a form to a branded investment memorandum landing in the inbox, at any hour, with nobody at a desk.

A property firm whose best leads arrive cold / Whose it was

A London property acquirer, live in production

A property acquisition and agency business working across London and the Home Counties, buying commercial, mixed-use and unbroken residential blocks. The system is embedded on their own site and signed off into production. What is not known is how many captured leads have become deals, because that figure has never been shared with us.

Read the full story See the appraisal tool

The problem

Case study: Property acquisition and agency

A firm whose best channel only works by hand / The problem

Your best channel is the one nobody scales

You already know the letters work. A planning decision lands, the owner is suddenly thinking hard about what happens next, and a letter that arrives that week gets answered. But somebody has to find the decision, work out who owns the building, write the right letter for approved or refused, print it and stamp it. So it happens in bursts, whenever somebody has a quiet week, and the rest go unwritten.

A firm whose best channel only works by hand / What we found

The letter was never the hard part

Mapping it end to end put the difficulty somewhere awkward. Thirty-eight council planning portals, each with its own layout, none of them agreeing on anything. And behind that, the question that decides whether any of it matters: who actually owns this building, and where do you post to. Get the owner wrong and every clever thing upstream is a wasted stamp. Meanwhile the pile of decisions nobody had written to had reached about five thousand.

A firm whose best channel only works by hand / What we built

Decision to doormat, without printing anything

A planning decision now surfaces on its own, gets tested against a tight list of what counts and what does not, resolves an owner and a postal address, and drafts the right letter for approved, refused or newly applied, on the existing letterhead. Nothing is written to twice. A second and third touch schedule themselves and stop the moment somebody responds. Nothing leaves without a tick first.

A firm whose best channel only works by hand / What changed

Built, proven, and not yet posted

This is the honest one. Not a single letter has gone in the post. The chain runs end to end on real planning decisions, the letters render correctly on the real letterhead at between 82p and about a pound each, and the whole thing sits in proof mode where it prints nothing, posts nothing and charges nothing. What is missing is not code. It is an owner-data subscription, a CRM token and credit on the print account. Roughly five thousand letters are queued behind that switch.

0 postedLetters actually in the post so far. Everything between a council planning decision and the doormat is built and proven end to end; the hold-up is three accounts, not code.

A firm whose best channel only works by hand / Whose it was

A London property firm, one switch from live

A property acquisition and agency business across London and the Home Counties, the same one behind the instant appraisal tool. The channel this automates already works when done by hand, at a return the owner puts at roughly six to one, though that is his own figure for his own manual mailings and nothing built here has measured it. Going live needs three accounts turned on, not more building.

Read the full story See the operations suite

The problem

Case study: Property investment and deal sourcing

Property deal sourcers and small portfolio investors / The problem

Hundreds of listings. Two of them are real.

Saturday morning, portal open, coffee going cold. You are scrolling because the deal that pays this month is buried somewhere in hundreds of listings, and the only way to find it is to look at all of them. By the time you have pulled rent comps for the four that looked promising, it is Sunday night and two are already under offer.

Property deal sourcers and small portfolio investors / What we found

The shortlist was shorter than anyone wanted to admit

Mapping how the numbers actually got made surfaced two things nobody wanted to hear. The same house was appearing three times under three different agents, so a shortlist of six was really two. And gross yield, the number everyone quotes, dies quietly on service charge: flats that looked stronger than houses on gross came out worse on net. The honest count was smaller and the ranking was upside down.

Property deal sourcers and small portfolio investors / What we built

The morning starts at the offer, not the search

The scan runs before you are awake, across five towns, and what lands is not a list of listings. It is a short set of properties with the bid already worked out: what to offer, what that offer actually nets after charges and running costs, and how far the seller's situation lets you push. Duplicates, retirement flats, shared ownership and cash-only stock never reach you.

Property deal sourcers and small portfolio investors / What changed

The best thing it found was the wrong town

It runs every morning on live listings. On the scan behind this, 127 confirmed properties across five towns, 83 of them still clearing 7 per cent net once the recommended offer, service charge and running costs come off, with no duplicate address anywhere. The most valuable output was not a deal though. It was proof that the town being worked hardest ranked third of five. No fee has been earned through it yet.

83 of 127Confirmed properties across five towns that still clear 7 per cent net once the recommended offer, service charge and running costs are taken off.

Property deal sourcers and small portfolio investors / Whose it was

A two-person sourcing venture, still deciding

Built for a part-time property sourcing venture in the East of England. Two people, aiming at one or two sourced deals a month. The proof was built and given away free, the engine has been running daily on live listings since July, and two ways to pay for the full build sit on the table. Neither has been signed. Nothing has completed through it.

Read the full story See the deal engine

The problem

Case study: Serviced accommodation and short-let management

A property business doing all its outreach by hand / The problem

You know who your customer is. Probably.

Empty nights are the whole problem. You have properties near an airport and a list of local firms that ought to need them, so you work it by hand: find a company, find a name, write something, send it, remember to chase. Four messages out, one reply, a week gone. Meanwhile the calendar has gaps in it next month and you are running the lot from another timezone.

A property business doing all its outreach by hand / What we found

Your target list was mostly a guess

The hand-worked list said construction firms near the airport, about ninety of them. It was a perfectly reasonable assumption and nobody had ever tested it, because at four messages a week there is no test to run. That is the uncomfortable part. You had no way of finding out you were wrong. Volume was never really about sending more, it was the only route to learning anything at all.

A property business doing all its outreach by hand / What we built

A hundred a day, and a scoreboard

Two engines now find companies and landlords, work out who to write to, draft in your own voice from your own templates, and hold everything behind an approval before a word leaves the building. Follow-ups run themselves once a first touch is approved, and replies come back sorted into interested, later, and no, so the inbox only holds the ones worth reading.

A property business doing all its outreach by hand / What changed

Construction was noise. Film was the answer.

Three genuine human replies in the first full week at proper volume, one of them warm. Not one came from the original list. All three came from film location agencies and travel, sectors added late. The construction firms, the ones worked by hand for months, returned out-of-office replies and bounces. No booking has come from it yet and no landlord has signed. What has come back is the answer to a question that could never previously be afforded.

3 repliesThree genuine human replies in the first full week at volume, and not one of them came from the sector the business had been working by hand for months.

A property business doing all its outreach by hand / Whose it was

A family short-let business, still in proof

A family-run serviced accommodation business with a small portfolio near an airport and a management arm it is trying to grow, run day to day from the other side of the world. The engines are live and sending. There is a target on the table for warm conversations by a date, with a refund attached if it is missed, and that clock is still running as this is written.

Read the full story See the outreach engine

The problem

Case study: Serviced accommodation and short-let management

Anyone onboarding clients before they have committed / The problem

You did the work before anyone said yes

Someone shows interest, so you send the long form. Property details, insurance, access, certificates, the lot. Then you chase it. Then you have the call, then you work out whether the place is even legal to let this way, then finally you talk about money. By the time anybody signs anything you have spent a fortnight on a property you might have to turn down. Some of them were never going to sign.

Anyone onboarding clients before they have committed / What we found

Three questions were asked, then quietly ignored

Walking it step by step turned up two things worth wincing at. The order was backwards: the deepest, most tedious intake ran before the agreement and the set-up fee, which is precisely the wrong way round for the person doing the work. And three questions already sitting on the form, about whether the mortgage or lease permits short lets, whether planning and licensing were checked, and whether it is legal to let this way, were being collected and then never acted on. Any one of them being a no should end the conversation. None of them did.

Anyone onboarding clients before they have committed / What we built

The order now enforces itself

Rebuilt as eleven steps in the sequence that protects you: qualify first, review as a team against a scored suitability card, send a written offer with a tier and a fee on it, take the signature and the set-up fee, and only then open the detailed intake. Nobody can reach the long form early, because the front door shows a different page depending on how far along the person actually is.

Anyone onboarding clients before they have committed / What changed

The mistake is impossible now, not discouraged

Nobody can be onboarded before signing and paying, because that route no longer exists. A permission answer that would sink the property blocks the offer at source instead of surfacing three weeks and several hours later. Every step shows what was actually submitted alongside what to do next, which was the thing being asked for. It is deployed, live, and walked end to end on test properties. No real landlord has yet come through it from start to finish.

3 questionsThree permission answers were being collected and then ignored. They now stop an offer being sent at all, before anybody spends a fortnight on a property that was never going to work.

Anyone onboarding clients before they have committed / Whose it was

Live, proven, waiting for its first landlord

This is the onboarding half of a family serviced accommodation business's management arm, the side that takes on other people's properties for a fee. It is deployed and in the team's hands. Two dependencies sit outside the build: a live e-signature provider and live card payment for the set-up fee, so today the money gate fires on a signal rather than a real signature. Until a landlord goes through it, the honest status is proven, not proved.

Read the full story See the onboarding flow

The problem

Case study: Serviced accommodation management

Serviced accommodation and property management operators / The problem

Miss one hand-off and the whole thing quietly stops

A landlord signs. Then someone has to notice. Then someone duplicates a template, sends an email, hops to the contract tool, hops back, makes a folder, chases a missing certificate, books a cleaner, an inventory, a photographer, and updates three databases that do not talk to each other. Five or six people touch it. Forget one step and nothing moves, and nobody finds out for days.

Serviced accommodation and property management operators / What we found

Only the physical steps actually needed a person

Walking it end to end produced the first complete written version of the process that had ever existed. Until then it lived in three places: a template, a screen recording, and one person's head. Counted honestly, step by step, only the photographer and the work that happens at the property needed a human at all. Everything else was people hand-carrying the same information between five tools.

Serviced accommodation and property management operators / What we built

It runs the process and asks before it sends

The whole onboarding now exists as something that runs rather than something people remember. Mechanical steps execute. Judgement steps arrive drafted, in the right person's name, waiting for one tap. A missing gas or electrical certificate gets spotted and the chase email writes itself with only the missing item on it. And you can ask where a property is up to without anyone stopping to compile an answer.

Serviced accommodation and property management operators / What changed

It has never touched the real business, deliberately

Twenty of the twenty-one steps ran unattended, end to end, on a structural replica built in separate accounts with invented data, every outbound message hard-forced into one sandbox inbox so it could not reach a real landlord. That was the rule from day one: rebuild the process somewhere safe rather than experiment on a live business. What exists is a proven process waiting on a decision, not a changed operation.

20 of 21Onboarding steps that ran unattended end to end on a structural replica, with every outbound message forced into a sandbox inbox.

Serviced accommodation and property management operators / Whose it was

A short-let manager, nine staff, still a replica

Built for a UK company that runs short-let and contractor accommodation on behalf of landlords: nine people plus outsourced partners for listings, pricing, inventory, photography and books. Nine departments and 59 workflows now run inside it, each carrying a tag saying whether it came from real documented procedure, real fragments, or standard practice still to be confirmed. It is live to look at, and not yet connected to anything real.

Read the full story See the company OS

The problem

Case study: Garden maintenance and landscaping

A one-van trades business run off a spreadsheet / The problem

Your whole business lives in one spreadsheet

Every customer, every price, every round day sits in one workbook you open on a phone in the van. You know roughly what comes in. You do not know which gardens pay for themselves, who is quietly underpriced, or what next March looks like. So you say yes to everything and work Saturdays to cover the gap.

A one-van trades business run off a spreadsheet / What we found

The wall was eight customers away

Rebuilt from the actual hours rather than the headline, the round came out 83 per cent full. Roughly eight more gardens and one pair of hands runs out of daylight, somewhere around £5,342 a month. The ten-grand-a-month goal was never a marketing problem, it was arithmetic. And more than a quarter of all the revenue ever sold had already churned back out again.

A one-van trades business run off a spreadsheet / What we built

A business you can interrogate, not just read

The spreadsheet became something you sign into. Click any figure and it opens up and tells you what it is made of, right down to the failed Direct Debit sitting behind it. Whoever is out on the round gets their own day view, tick targets big enough for wet hands, and photographs a receipt that lands for approval and goes out with that week's wages. Where the data is thin, the screen says so instead of flattering you.

A one-van trades business run off a spreadsheet / What changed

The ceiling stopped being an opinion

The wall is still there. What changed is that it now has a date on it, March 2027, and a number underneath it. Hire, reprice, or lean harder into project work: all of those used to be gut calls made in a van, and are now arguments with figures attached and caveats stated. Direct Debit collection is built and waiting on one API token to switch on.

£5,342The monthly point at which one pair of hands runs out of daylight. It did not appear anywhere in the spreadsheet the business was being run from.

A one-van trades business run off a spreadsheet / Whose it was

A one-gardener round, 39 customers, one van

A garden maintenance subscription business in the north of England: 39 customers, £4,413 a month, one owner and one employee. Phase 1 is live and in daily use, in its own brand colours. The capacity ceiling rests on only 19 of 39 visits having a real time logged, and the app labels it that way rather than hiding it. Phase 2, the part that answers enquiries without the owner coming off a ladder, is quoted and not yet built.

Read the full story See the company dashboard

The problem

Case study: Construction and industrial recruitment

Agency owners running a dialling floor / The problem

Your floor burns the list by Wednesday

You bought the leads. By Wednesday morning the good ones are gone and your callers are working the leftovers, ringing companies that dissolved two years ago and supermarkets that somebody somewhere tagged as construction. The dials still happen. The requirements stop coming in. And you genuinely cannot tell whether the problem is the list, the script, or the people holding the phone.

Agency owners running a dialling floor / What we found

Half the list had been invented

Map it end to end and the scraper gives you the answer, and it is worse than a bad supplier. Whenever a job board blocked it, it invented companies rather than returning none. Of 512 records, 449 were fabricated outright. Strip the dissolved firms, the wrong sectors and the rival agencies as well and a few dozen usable leads survive. Your floor was not failing to convert. It was ringing fiction.

Agency owners running a dialling floor / What we built

A floor that only rings real companies

The fallback that invented companies was deleted outright, so a blocked job board now returns nothing instead of fiction. Nothing reaches one of your callers without being an active company at Companies House, in construction, hiring right now, with a real human being to ask for. Your morning queue arrives ranked by how fresh the vacancy is, so the hottest hundred get called first.

Agency owners running a dialling floor / What changed

224 firms on the board, all real

A later rebuilt pull put 224 construction firms on the board, every one active at Companies House, none invented, none dissolved, no guessed phone numbers. Two thirds carry a named decision maker with a verified email. The rest carry a real director's name and the company switchboard. Honest weekly yield is 150 to 250, not the 500 originally scoped. No placement has come through it yet.

449 deletedInvented and dead records removed from a lead database a sales floor had been dialling as though it were real.

Agency owners running a dialling floor / Whose it was

A construction desk, and not an arm's-length one

A UK construction and industrial recruitment agency with a dialling team, and the engine runs on their own server feeding their own floor every morning. Worth saying plainly: this desk belongs to a business partner, so it is not a customer who chose us from a shortlist. The purge and the rebuild are done and verified. No placement has been traced back to it. Until one is, this is a supply fix and an honesty fix, not a revenue claim.

Read the full story See the lead engine

The problem

Case study: Construction recruitment, permanent desk

Recruiters sitting on a dormant candidate database / The problem

Ten years of candidates, nobody speaking to them

You have tens of thousands of candidates and no time to market a single one. Business development is what happens in the hours left over, and there are no hours left over. So new business leans on the same three accounts it leaned on last year, while the database you spent a decade building sits there earning nothing, because working it properly costs exactly the time you do not have.

Recruiters sitting on a dormant candidate database / What we found

Everything worked right up to the reply

Map it end to end and the sophistication turns out to be in the wrong half. Scoring, tiering, fatigue caps, drip windows, editable approvals, all upstream. Downstream of a reply, where every pound you earn actually happens, six empty columns and nowhere in the database to store a fee. Two genuine warm replies had sat untouched for over a fortnight, because no phone number was ever captured.

Recruiters sitting on a dormant candidate database / What we built

Outreach that costs nobody an hour

Every week it reads a batch of your dormant candidates, works out which firms would genuinely hire them, finds the person who does the hiring, and writes the approach in your voice with a reason attached you would be happy to read out on a call. Nothing sends until you click approve. It is not built to close a placement. It is built to open a conversation.

Recruiters sitting on a dormant candidate database / What changed

Four replies. No placements. The audit says why.

The honest numbers off the live engine: 197 candidates, 8,283 contacts enriched, 341 emails sent, four replies, zero placements recorded. Two of those four were an auto-responder and a complaint about where the details came from, so the true warm rate is nearer 0.6 per cent. Almost 8,000 enriched contacts were never sent to at all. Everything upstream of the reply works. Everything downstream of it was never built.

0Placements recorded after 197 candidates, 8,283 contacts enriched and 341 emails sent, on the engine's own internal audit.

Recruiters sitting on a dormant candidate database / Whose it was

Our own venture, with nothing to show yet

A UK construction recruitment desk running the first and only live instance, and it is only fair to say this one is ours: a venture we half own, not a client who paid us. Its own four-agent audit found six empty columns after the reply, a fee storable nowhere in the schema, and zero placements. This panel is here because it is true, not because it sells, and the reframe it produced is the useful part: the engine starts conversations, it does not close placements.

Read the full story See the recruitment engine

The problem

Case study: Technology recruitment

Tech recruitment founders with a large idle database / The problem

Ten people, no hours, no new clients

Every new client conversation costs a consultant an hour they have not got. Ten people, all already filling roles, and business development is the first thing dropped when a day gets busy. Meanwhile 150,000 candidates sit in your system going quiet, because working them properly takes exactly the hours nobody has. It is not a data problem. It is a capacity problem, and no tool has ever sold you capacity.

Tech recruitment founders with a large idle database / What we found

Tech does not tier the way construction does

Map how matching actually has to work here and the construction model falls apart. There, a firm one rung below a candidate's current employer is a defensible target. In tech it means nothing. Your back-end developer at a bank belongs at a council or a charity startup if the stack lines up, and nowhere near another bank if it does not. Match on career path and you will produce confident, useless lists at speed.

Tech recruitment founders with a large idle database / What we built

A reason you would read out on a call

A working slice of the engine pointed at real candidates, not a slide deck. It reads a person, finds firms that would genuinely hire them, names who does the hiring, and attaches one line explaining why that company. Weak matches are left visible instead of quietly dropped, so you can see where it is wrong before you decide whether to trust it. Nothing leaves the building unapproved.

Tech recruitment founders with a large idle database / What changed

Sceptical at the demo, pilot agreed five days later

Twenty real tech candidates produced 47 matches, every target verified as an active UK company, each with a named decision maker and a written reason averaging 4.2 out of 5. Deliberately poor matches were left in on the day rather than hidden. Fifty-one minutes turned a room that expected none of it to work into one leaning towards buying. Five days later a paid pilot was agreed.

5 daysFrom a sceptical founder watching a live demo to an agreed paid pilot carrying a full money-back guarantee.

Tech recruitment founders with a large idle database / Whose it was

Pilot agreed, and nothing proven yet

A 20-year-old UK technology recruitment firm, ten staff, one CRM carrying everything. At the end of August they agreed a four-week paid pilot with a full refund if it fails to produce even one genuine new client conversation. It has not gone live and the money is not banked. And the single thing they said still needs proving, matching by technology stack rather than career path, is precisely what the pilot has to prove.

Read the full story See the tech recruitment build

The problem

Case study: Hospitality finance recruitment

Niche recruiters where everyone knows everyone / The problem

One wrong approach and the account is gone

Your market is small enough that everybody owns everybody. Send a financial controller's CV to a hotel group and you can discover you have just marketed your own client's employee back to their parent company, trading under a different name. One of those loses an account you spent years building. Which is why nobody on your desk will let outbound run without reading every line of it first.

Niche recruiters where everyone knows everyone / What we found

The risk sits above the message

Map it and the danger is a layer higher than anyone was looking. The names do not match. A property, its management company and its ultimate owner can trade under three unrelated brands, and nothing on a CV tells you which group somebody actually sits inside. Check the hotel name alone and you are safe every single time, right up until the one day you are not.

Niche recruiters where everyone knows everyone / What we built

It refuses to pitch a candidate's own owner

Before a word is written, every candidate is resolved to the group that genuinely owns their property, and every target is checked against it. Sister properties trading under other names come off the list before any outreach is drafted, and the excluded group is shown on screen, so you can see exactly what it declined to do and why. The check runs whether anyone remembers it or not.

Niche recruiters where everyone knows everyone / What changed

It caught what the name would not

On the test run, a financial controller at a five-star London property had two well-known groups pulled from her target list, because both quietly sit under the same owner as her current employer despite carrying entirely different names. Four suppressions fired across the run. Several candidates honestly returned no matches at all, and that was left on screen rather than padded out.

4 suppressedTarget companies removed from candidates' lists because they sat under the same ultimate owner as the candidate's current employer, despite trading under different names.

Niche recruiters where everyone knows everyone / Whose it was

It works. Nobody has ever used it.

A 40-year-old UK hospitality finance recruitment firm, and the honest status is that this is a sales demo, not a delivery. It runs on sixteen candidate profiles we wrote ourselves. No consultant has ever touched it, nothing has been bought, and no relationship has ever been protected by it. The ownership check genuinely works. It has simply never been pointed at anything real.

Read the full story See the hospitality build

The problem

Case study: Recruitment

Agency owners running an outbound calling team / The problem

Four callers, one spreadsheet, nobody knows who rang who

You bought the list. You hired the callers. Now three of them are working the same tab of the same spreadsheet, one has just rung a company another spoke to on Tuesday, and the callback someone promised for Thursday exists only in a WhatsApp message. At the end of the week you can tell how many calls were made. You cannot tell what happened on any of them.

Agency owners running an outbound calling team / What we found

The list was never the problem. Ownership was.

Mapping how a call actually gets worked showed the list was the easy part. Everything that matters happens after the call: who owns this lead now, what was said, when the callback is, and who else can see it. A hard review before launch also found the first build let any caller open any lead just by typing its number. On a floor of people calling your best prospects, that is not small.

Agency owners running an outbound calling team / What we built

Every call finds its own way to the right lead

Each caller now opens a list that is theirs, plus an unclaimed pool, and cannot reach anyone else's. Dialling still happens in the phone system, the way it always did, and the call comes back on its own and attaches itself to the lead with the talk time and the recording. Callbacks and tasks surface on one Today screen. You see the whole floor without asking anyone for a number.

Agency owners running an outbound calling team / What changed

Live since August, and one feature had to go

It has been live since August, with 140 calls logged automatically against the right lead at the last verified check, recordings attached, nothing typed in by hand. One feature was deliberately removed a week in: the in-app call button placed a two-legged call that degraded the audio, so it went back to a plain dial link. A worse-looking product, better-sounding calls. No placement or revenue figure exists yet.

140Calls logged automatically against the right lead, with talk time and recording attached, at the last verified check.

Agency owners running an outbound calling team / Whose it was

A recruitment group, mid outbound build-out

Built in August 2026 for a UK recruitment group standing up a small outbound calling floor: two admins, a view-only manager for oversight, and diallers working assigned leads plus a shared pool. It went in seeded with 318 prospects, which is a list to call rather than a result. It is live and in use. What it is meant to produce has not been counted yet.

Read the full story See the dialler CRM

The problem

Case study: Professional services mentoring, architecture

Mentors and coaches selling to micro-practices / The problem

The list you need cannot be bought

You sell to one-person practices, and there is no field anywhere that says one person. Every list you buy arrives as a mix of firms too big to need you and firms that folded in 2019. You pay for it twice: once at purchase, then again in the evenings somebody spends checking rows, before a single message has gone anywhere.

Mentors and coaches selling to micro-practices / What we found

Two thirds of your best market is unmailable

Map the market and the uncomfortable half arrives quickly. Sole traders count as individuals under UK marketing law, which means a large share of the practices that fit you perfectly cannot lawfully be cold-emailed at all. Of 2,595 that matched on the work they publicly advertise, 1,682 fall the wrong side of that line. Most list vendors will happily sell you those rows anyway.

Mentors and coaches selling to micro-practices / What we built

It will not let you break the rule

Each practice's own website is read for the work it actually advertises, and each row is tagged with whether emailing it is lawful. Then the part that matters to you: try to log an email against a practice that is not allowed to receive one and it refuses, and tells you which reason applies. Fifty a day, capped, so nobody can panic-blast on a slow week.

Mentors and coaches selling to micro-practices / What changed

A defensible list, and almost nothing sent

2,595 practices that genuinely advertise the work, 855 lawful to approach by email, and 1,917 carrying a phone number. 1,266 of those phones sit on rows email is blocked from, and phone is a different legal test, so the weakest half of the list becomes the workable half. What has actually happened since: two logged touches. The list is real. The outreach never started.

1,682 blockedPerfectly qualified practices the system refuses to let anybody cold-email, because UK law counts sole traders as individuals.

Mentors and coaches selling to micro-practices / Whose it was

Unsold, unrun, and one permission short

Built for one UK mentor selling to architecture practices, and the honest status is that nobody has bought it and nothing has been sent. It is missing a working AI key and the client's real customer examples. There is also a permission still to obtain over how the underlying professional register may be used commercially, and until that is in writing this one stays a story about rules, not a campaign.

Read the full story See the outreach engine

The problem

Case study: Planning and architecture consultancy, care sector

Specialist consultants whose expertise never gets published / The problem

You know the expensive mistake. You never post it.

Someone is about to buy a building that will never become the thing they want it to be. You could save them in three paragraphs. You have been meaning to write it for a year. Posting loses to fee work every single week, and it should, so the person best placed to be the voice in the niche stays silent and the wrong building gets bought.

Specialist consultants whose expertise never gets published / What we found

The bottleneck was not writing. It was choosing.

The uncomfortable finding was that the obvious fix makes it worse. In a narrow regulated specialism, a fluent generic post tells buyers you are any consultancy. Generic AI is worse still: confident, and perfectly willing to invent a case study or imply a planning outcome is certain, under your name and your professional standing. The posts worth writing are the ones a cautious competitor would not, and those take the longest.

Specialist consultants whose expertise never gets published / What we built

The reviewer never sees the post being written

Your positioning, your five buyer types, what each of them is actually afraid of, and the specific things you know that give you standing to speak, all live in one file you can edit. One pass decides only what is worth saying and to whom, and never writes a word. Another writes it. A third reads the finished post cold, scores how specific it is, and can kill it outright.

Specialist consultants whose expertise never gets published / What changed

Five posts, graded 8.4, none published yet

Five posts exist, written and then graded by a reviewer that never saw them being drafted. They scored 8, 9, 8, 8 and 9 out of ten for specificity, and one was flagged for a policy check before a human had even read it, which is the system doing exactly what it was built to do. Nothing has been published. There is no reach or enquiry data because nothing has gone out.

8.4 / 10Average specificity score across the five posts, graded by a separate pass that never saw them being written.

Specialist consultants whose expertise never gets published / Whose it was

A specialist consultancy, waiting on one key

Built in a single day for a founder-led planning and architecture consultancy in the care sector, whose buyers make six and seven figure decisions on buildings. It is deployed behind the firm's own site, login protected, with five reviewed posts sitting on the board. It is one credential away from generating more, and the planning specifics still need the founder's own sign-off.

Read the full story See the content engine

The problem

One hour. Every process in your business mapped.

No charge and no deck. You leave with a ranked list of what can be automated, and a number for what it is costing you to keep doing it by hand.

Book a call

MAX.AI · 2026 · Bespoke AI systems · United Kingdom