Talk numbers
Work

Client stories: the question each business couldn't answer, what we built and what changed.

All work
Built

Tools we built, for clients and for ourselves, when nothing we could buy did the job.

Open to anyone

Runs our work

For clients

See all builds
AI

AI moves fast. Here's how we use it in our own work, and where to start in a business your size.

The AI hub
Approach

Every business is different, so we always start with a conversation.

How we work
  1. 15 minA quick callYour goals, your numbers, and whether we're a good fit.
  2. 1 hourDiscoveryWhere you're losing enquiries, and what's already working.
  3. ThenA clear planPaid ads, a custom build or both, as a one-off project or ongoing.

What we connect

About

We're a small London team of marketers, and we use AI to go faster.

About us
  • Cam BowenCam BowenCo-founder
  • Ben PowerBen PowerCo-founder
  • ElectraElectra
  • MahirMahir
  • NatashaNatasha
  • RahabaRahaba
  • ZobaerZobaer
AI by company size

Pick your company size and we'll show you where to start.

Most AI advice ignores how big you are. A business of eight people and one of eight thousand don't have the same problems, the same risks or the same first move. For each size you'll see what companies like yours are already doing, what to do this month, what to leave alone for now, and three questions to ask anyone who offers to help (us included).

Talk numbers How we use AI
Pick a size

Where companies your size are starting

Pick the size closest to yours. We review each guide every quarter, and the date's on it.

Pick your size

We go by headcount: Small is under 20 people, Medium is 20 to 200 and Large is 200 and over. If you're near the line, read both.

Small · Under 20 people

Where to start with AI when you have under 20 people

  1. Would it terrify you to know your competitor was answering every enquiry in minutes, at 3am, while yours wait until Monday?
  2. Would it terrify you to know your competitor had a month of content written before lunch?
  3. Would it terrify you to know your competitor could tell exactly which ads make money, and yours are still a guess?

A government survey found 14% of micro businesses (fewer than 10 staff) were using AI. The most common barrier businesses cited, across all sizes, was that they hadn't identified a use for it (71%) (DSIT, fieldwork February to May 2025). So start with one job.

Source: Department for Science, Innovation and Technology, "AI Adoption Research", updated 13 February 2026, fieldwork 12 February to 2 May 2025.

What companies your size are already doing

  1. Effort Days

    Answer enquiries in minutes, 24/7

    Every new enquiry gets a sensible first reply and goes straight into your CRM, and you get an alert with the details.

    Stays human Anything that's a quote, a price or a promise, you write or approve yourself.

    ProofHow we run form to CRM to alert

  2. Effort Hours to days

    A month of content in a day

    In one sitting you talk through your week, AI drafts a month of posts and emails, and you cut it down to what sounds like you.

    Stays human You choose what to say and do the final read. Nothing goes out without you reading it.

    ProofA big batch of FAQs drafted and loaded in one sitting

  3. Effort Days to weeks (it starts with tracking)

    Know which ads actually make money

    Your ad spend is tied to real leads and sales, so you can see what paid for itself and stop what didn't.

    Stays human You decide what to cut and what to spend more on.

    ProofAds priced against what a client is worthOne person's trail, not three GA4 users

  4. Effort Days

    Build your own little tools

    A quote calculator, a booking checker or a small internal app, made with Cursor or Claude Code, for the one job you keep doing by hand.

    Stays human You decide what it should do and test it with real cases. We read every change before it runs.

    ProofMarkdown ViewerImage CompressorCursor

Show 4 more
  1. Effort Hours

    Voice note to a finished post or email

    After a job you record two minutes on your phone, and AI turns it into a post, a case note or a customer email that sounds like you.

    Stays human You say what actually happened and approve the wording.

  2. Effort Days

    Quotes and invoices chased for you

    Each week, overdue invoices and unanswered quotes get a polite reminder drafted for you. You send it or click approve.

    Stays human You decide who gets chased and how firmly. Nothing's sent without you.

    On our own work Our own rule: nothing gets sent unless a person asks.

  3. Effort Hours

    Reviews asked for and answered

    After each job, a short review request goes out, and replies to reviews are drafted for you to edit.

    Stays human Anything about a complaint or a particular person, you write yourself.

  4. Effort Days

    A weekly "is my site working?" check

    An automatic check opens your forms, tests your tracking and tells you on Monday if a form or a tag has broken.

    Stays human You fix it or ask someone to.

    ProofTracking that proves a purchase Our Site Watch bot watches live sites for problems and files what it finds.

First step this month

Pick the one job that eats most of your evenings. For most small businesses, that's replying to enquiries. Write down the five questions you get asked most and your real answers, then set up one place where every enquiry lands (a form that feeds a simple CRM and sends you an alert). Don't add AI to the replies yet. Once every enquiry lands in one place, add a drafted first reply that you approve.

Not yet

  • Don't buy five tools. Pick one general assistant and one automation tool.
  • Don't let AI send anything to a customer before you've read it.
  • Don't paste customer data into a free tool until you've read what it does with it.
Three questions for anyone who offers to help, us included
  1. What exactly will you do with my customers' data, and where does it go?
  2. Who checks the work before it reaches a customer, and how?
  3. If a tool you use changes or disappears, what do I still have?

Read the full guide: under 20 peopleWant us to set it up for you?
Medium · 20 to 200 people

Where to start with AI when you have 20 to 200 people

  1. Would it terrify you to know your competitor's sales calls were summarised, scored and in the CRM before the rep had put the kettle on?
  2. Would it terrify you to know your competitor's monthly report wrote itself, and their team spent that week on the work instead?
  3. Would it terrify you to know your staff are already using AI at work, with no shared rules, and your competitor has written theirs down?

Among firms that reported barriers, the ONS found a lack of expertise was one of the most common, and about 18% of firms with 100 to 249 staff said it had held them back from using AI (ONS, 20 July 2026). That's why rules and ownership come first.

Source: Office for National Statistics, "Artificial intelligence in UK businesses: 2023 to 2026", 20 July 2026.

What companies your size are already doing

  1. Effort Days

    Every sales call summarised into the CRM

    The call's recorded and transcribed, and a summary, the actions and the next step land on the right record.

    Stays human The follow-up message. A person writes it.

    ProofFirefliesTwenty Our Meeting Intel bot does this for us: it reads each meeting, writes notes and tasks, and doesn't send anything.

  2. Effort Weeks

    Reporting that writes itself

    The numbers get pulled together and a first draft of the commentary is written. Then a person adds the "so what".

    Stays human A person decides what the numbers mean and what to do next.

    ProofOngoing reportingMeasurement treated as the product

  3. Effort Days

    Shared AI rules for the whole team

    One short written rulebook: which tools are approved, what data never goes in and who checks what. It's loaded into the tools themselves, so it gets followed instead of filed away.

    Stays human Agreeing the rules, and owning the exceptions.

    ProofHow we run one rule set across two AI tools

  4. Effort Days to weeks

    Automations replacing admin busywork

    n8n and AI handle data entry, sorting and handoffs: a form comes in, a record's created, a task is set and the right person gets an alert.

    Stays human Anything that needs judgement goes to a person. So do errors, instead of disappearing.

    ProofOur own forms-to-CRM flowTwenty and n8n

Show 4 more
  1. Effort Days

    First-draft proposals from the call

    After a discovery call, a proposal outline is drafted from the transcript and your past work, ready for a person to rewrite.

    Stays human Scope, price and promises.

    ProofFireflies

  2. Effort Weeks

    "Ask our docs" for your own team

    One searchable place where your staff can ask how a process works and get an answer with a link to the source.

    Stays human Keeping the docs accurate. Every page has an owner.

    ProofOur own stack docs, searchable, with an Ask AI index

  3. Effort Days

    Launch checks before every campaign

    Before any money's spent, your tracking, consent settings and forms are checked, and anything broken gets listed.

    Stays human Signing off the launch.

    ProofTracking that proves a purchase

  4. Effort Days to weeks

    Lead triage to the right owner

    New leads are read, tagged by type and urgency, and routed to the right person with a short brief.

    Stays human The relationship. Anything unusual goes to a person first.

    Proofn8nSeparating audiences so each gets the right journey

First step this month

Write your one page of rules before you buy anything. List the tools people already use, what data must never go into them, and who checks what. Then pick one job you repeat (usually getting call notes into the CRM) and make it work for one team before you roll it out. If your tracking and CRM data are messy, fix that first, because AI on bad data gives you confident wrong answers.

Not yet

  • Don't roll out an AI tool across the company until you've got rules and an owner.
  • Don't build agents that act on their own (sending, spending, publishing) until your repeated jobs run well with a person checking.
  • Don't judge success by how many people logged in. Measure the one job: time saved, errors, how fast you reply.
Three questions for anyone who offers to help, us included
  1. Which of our systems will this connect to, and what can it change without asking a person?
  2. How will we know when it's wrong, and who gets told?
  3. When you leave, what do we keep: the rules, the workflows, the logins, the documentation?
Large · 200 people or more

Where to start with AI when you have 200 people or more

  1. Would it terrify you to know your competitor had an approved AI stack, with data boundaries, running across every team, while you have forty tools and no list?
  2. Would it terrify you to know your competitor's pilot shipped in weeks, while yours is still in review?
  3. Would it terrify you to know that when a customer asks an AI assistant who to buy from, it names your competitor and not you?

A government survey found 36% of large businesses (250 or more staff) were using AI (DSIT, fieldwork February to May 2025). Using AI is one job. Having proper rules around it is another.

Source: Department for Science, Innovation and Technology, "AI Adoption Research", updated 13 February 2026, fieldwork 12 February to 2 May 2025.

What companies your size are already doing

  1. Effort Weeks

    An approved AI stack with data boundaries

    A short list of approved tools, a plain rule for what data goes where, and one owner. Your staff know what to use and what never to paste in.

    Stays human Approving tools and exceptions. A named person is accountable.

    ProofOur own rules: hooks that stop one tool doing another's job, and nothing sent unless a person asks

  2. Effort Weeks

    Agents watching your sites and tracking overnight

    Agents check pages, tags and forms while you sleep, and leave a short note or an issue for someone to look at in the morning.

    Stays human Anything that changes a live site. Agents suggest the change and a person merges it.

    ProofBrowser checks Our own bots watch like this: Site Watch watches live sites and files what it finds, and Ship Guard watches our repos and files issues. A person merges every fix.

  3. Effort Weeks

    Pilots that ship in weeks, not quarters

    One workflow, one owner, a written test for success and a review date. If it works, it goes live. If it doesn't, it stops.

    Stays human Deciding go or no-go, and setting the success test.

    ProofTechStack AnalyserUser Journey Analyser Both started small: TechStack for our own pre-pitch scans, Journey for a single client first.

  4. Effort Weeks to months

    Answer engines recommending you (GEO)

    Your content and key facts set out so ChatGPT, Gemini and similar tools can find them and repeat them correctly.

    Stays human Deciding what you want to be known for, and checking what the AI tools actually say about you.

    ProofContent and SEO

Show 4 more
  1. Effort Days to weeks

    A no-blame amnesty on unofficial AI, and a register

    Ask staff what they already use, with no blame, then record it all in one register: the tool, the data it touches, who owns it and what you've decided.

    Stays human Deciding what happens to each tool.

  2. Effort Weeks

    A test set before anything customer-facing

    Before an assistant talks to customers, run it against 50 to 100 real questions and have people score the answers. Run it again every time something changes.

    Stays human Scoring the answers, and deciding whether to launch.

  3. Effort Weeks

    Tracking and consent checked across every site and market

    Agents check tags, consent settings and where data goes across all your sites and markets, and flag anything that's drifted.

    Stays human Legal and privacy sign-off.

    ProofTechnical trackingMeasurement treated as the product

  4. Effort Weeks

    Brand and tone rules built into the tools

    Your brand rules, claims and banned phrases are loaded into the tools, so drafts start on-brand and fewer need rewriting.

    Stays human Approving claims and anything regulated.

    On our own work We do this on this site: one written rule set that every AI tool reads.

First step this month

Find out what's already in use. Run a no-blame survey of the tools people use and the data they touch, and write it up as a one-page register. You can't set limits on tools you haven't listed. Then pick one pilot, give it one owner, and write down the test before it starts.

Not yet

  • Don't sign a multi-year, company-wide AI contract before you've run a pilot.
  • Don't let an agent send, spend or publish without a person checking.
  • Don't run ten pilots at once. Two with owners will beat ten without.
Three questions for anyone who offers to help, us included
  1. Where does our data go, who at your end can see it, and is it used to train anything?
  2. Show us the review step. Who signs off, what gets logged, and can we audit it?
  3. What happens to the work, the prompts and the setup if we end the contract?
A note from us

We know this is confusing.

New tools and models turn up nearly every week, most advice is out of date within a month, and a lot of it's written by people trying to sell you something. We use AI on our own work every day, and we still have to work at keeping up.

So we've kept these guides short, dated and specific, and we'll tell you when we're guessing. If something here stops being true, we'll change it and say so.

Ben and Cam

Choosing help

Three questions for anyone who offers to help.

Ask us them too.

  1. What exactly will you do with our data, and where does it go?
  2. Who checks the work before it reaches a customer or goes live, and how will we know when it's wrong?
  3. If a tool you use changes, or we stop working together, what do we keep: the rules, the workflows, the logins, the documentation?

Each size guide above has its own version of these, for what matters most at that size.

Next step

Bring your numbers.

We'll tell you if the spend makes sense.

  1. A 15-minute callYour goals, your challenges, and whether we're the right fit.
  2. A 1-hour discovery callWe dig into your numbers. Worth a hundred emails.
  3. A concrete planAds, a custom build or both, as a one-off project or ongoing.
Talk numbers
15 minutes · No pitch deck · A straight answer