50% off pre-order$497 $24994 of 100 leftPre-order now
Work with me · Custom AI SDR builds

Stop renting an AI SDR. Own one built for your business.

Every way of fixing outbound right now is either built for everyone, distracted by other clients, or a job your reps cannot do on top of their quota. So I build you the thing all three are trying to be: a custom AI SDR trained on your business, running email and LinkedIn, proving exactly which meetings it produced.

You own the system, the data and the playbook. It is also the cheapest of the four to run.

Book a call

Start with a one-day workshop at $1,000, or talk through the full build. Two engagements at a time.

Teams I've built this function with
BowimiChannel CrawlerEvergon LabsNousScaletopia3am ThoughtsLenkieAbacumSynthesiaHeySarahLaurel AI BowimiChannel CrawlerEvergon LabsNousScaletopia3am ThoughtsLenkieAbacumSynthesiaHeySarahLaurel AI

Right now you have three options, and all three of them break.

Option 01

Turn your reps into GTM engineers

GTM engineering is a real discipline and it is genuinely hard. Your best rep is not going to become one on top of carrying a number. Worse, the title got hot, so the market filled with people who will tell you they can implement Clay and leave you with a half-built system nobody in the building can maintain. The tools are priced for the hype rather than the outcome.

Option 02

Buy an AI SDR

Built for everyone, which is the same as built for no one. It can only write what a settings form lets it know about you, so your outbound turns to slop. The people who reply to slop are the people worth least to you. That is the real cost: not the wasted send, but the pipeline you fill with the wrong logos.

Option 03

Hire an agency

You are one account among many, and your campaigns get whatever attention is left after the clients bigger than you. Six to eight thousand a month, forever, for a black box. When the relationship ends they leave with the playbook, the list logic and the copy library. You keep the invoices.

There is a fourth option. Build it once, own it outright, and run it for less per month than any of the three above. A custom AI SDR does what all three of those are supposed to do, at the standard all three promise and none of them hit.

Why generic AI SDRs write generic copy

They are one system serving thousands of companies through a settings form. The form has room for your company name and a tone slider. It has no room for the eight angles that died in the field last quarter, the exact phrase your best customer used to describe their pain, or the reason a company calling itself a platform is actually an agency you should never email.

I build that in. It is the difference between a tool configured with your business and a system built on it.

What actually gets built

An ICP library with definitions precise enough that an agent can judge a company against them, and reject the lookalikes that job titles and keywords cannot separate.
A persona library where each persona carries its own pain in the buyer’s words, hook angle, mechanism lines, proof points with real outcomes, CTA style, and banned framings: angles that died in the field and must never be used again.
An offer library of distinct angles, so your campaigns test positioning rather than subject lines.
A QC agent that visits every company’s website and verifies it against your definition before anyone gets emailed. This is what keeps the wrong logos out of your pipeline.
A copy agent writing from that material, not from a prompt template.
A revenue truth layer wiring calendar, CRM and sending platform together, so every meeting carries its real source and the evidence behind it.

That is a knowledge base the system reasons over. It is why the copy sounds like someone who did their homework, and why a tool serving a thousand companies through a form structurally cannot produce it.

Three layers, one system

Layer 01

Revenue Truth

Verified attribution, CRM hygiene, real pipeline value, data-quality watchdogs, Slack digests. The layer that makes every other number defensible.

Layer 02

Call-to-Cash

Every call transcribed, analysed, and turned into a CRM deal valued from what was actually said. Briefs before, follow-ups and post-mortems after.

Layer 03

The AI SDR

ICP → offers → list build → triage → website QC → copy → A/B → live campaign, across email and LinkedIn.

An AI SDR that can prove what it produced is a different product from one that cannot.

The same four options, priced

Rep as GTM engineer
$2–4k/moin tools, plus the headcount
  • Tools priced for the hype
  • Half-built, nobody can maintain it
  • Your best rep off their number
  • You own a system that does not work
Generic AI SDR
$1–2k/moforever
  • Generic by design
  • No attribution
  • You never own it
  • Fills pipeline with your worst fit
An agency
$6–8k/moforever
  • You are one client of many
  • Attribution is their word for it
  • They leave with the playbook
  • Their mistakes, on your list
Abbas Somji
Abbas Somji
CRO · InboxKit · Founder · Making Plays

UK's youngest head of sales at 24. Built the cold email engines behind 50+ SaaS and tech brands through my agency, then exited it at 28. Now CRO at InboxKit, scaling to $100m ARR. The first person this system was built for was me.

Abbas set up the cold emailing function for 3 different companies I was advising as VP. He became my go-to source to bring in when building a function from scratch.
Hassan TariqHassan TariqVP Commercial, Evergon Labs
Abbas eliminated the need for us to hire SDRs. Instead we hired AEs, as his cold email engine delivered 20 meetings per week on autopilot.
Jon RudoeJon RudoeChief Commercial Officer, Nous
Abbas trained our first sales hires, which helped us scale, and set the model for our AE and BDR team ever since. £200K ARR added in six months.
Dom BowcockDom BowcockCo-Founder, Bowimi

Three ways to start

Most people start at the top and move down the list as the evidence stacks up. You are not committing to the next rung by taking the first one.

Start hereThe workshop
$1,000 one day

A working session, not a slide deck. Comes off the build fee if you go ahead.

  • We map your ICP, personas and offer angles live
  • You leave with the ICP definitions written down
  • The keyword logic that decides who gets contacted
  • Where your current outbound is losing money
  • An honest read on whether a build is worth it for you

Best for: Finding out whether any of this fits before spending real money.

Most popularThe MVP build
$10,000 2–3 weeks

One ICP, one channel, live. The machine running on your business, at small scale.

  • One ICP defined precisely enough for an agent to judge against
  • One persona and one offer angle, built from your material
  • QC agent calibrated on companies you hand-label yourself
  • Copy written from that library, not from a template
  • One email campaign live and sending
  • Yours to keep and run, whatever you decide next

Best for: Seeing it work on your business before committing to the full build.

The 90-day build
$25,000 90 days

The complete system, both channels, handed over to your team.

  • Everything in the MVP, across your full ICP and persona set
  • Email and LinkedIn, both live
  • Revenue attribution wired through, meeting by meeting
  • Call intelligence: transcripts to CRM deals with real values
  • Your repo, your infrastructure, your keys
  • Team training and a documented playbook
  • Two ICP calibration rounds built into the scope

Best for: The whole machine, owned outright, running without me.

90-day build on a stack that needs a new CRM adapter$35,000
Optional: I keep running it after handover$3,500/mo
Optional: maintenance licence (vendor API changes, model updates, support)$1,000/mo
Your own tool costs, paid direct to the vendors~$470–2,400/mo

There is also a performance option on the full build: a lower fee against $150 to $300 per verified cold-email meeting. Verified means something specific and auditable here, which is why I can price that way.

Just want the thinking, not the build?

I also work with a small number of teams fractionally on go-to-market: pipeline strategy, outbound diagnosis, ICP and positioning work, or sitting in as a second opinion while your team builds it themselves. Hourly or on a retainer, depending on what you need.

Pricing on request. Tell me the problem on the call and I will tell you what it takes.

Discuss consultancy

What the 90 days actually look like

Ninety days, not sixty. A three-email sequence takes about three weeks to play out, then replies lag, then meetings lag. Sixty days buys one incomplete cycle and no evidence.

Days 1–14

Foundations

Credentials, ICP and persona libraries, offer angles, tier criteria. A full CRM schema and default-value sweep so the numbers mean something later.

Delivered: The libraries the whole system reasons over, written down and yours.

Days 15–45

Build

QC agent calibrated against companies you hand-label. Copy agent writing from your material. Funnel tuning against the first real sends.

Delivered: Your first email campaigns live, on a list where every company was verified.

Days 46–75

Expand

LinkedIn live. Cycle two. Copy and targeting tuned against real replies. Call intelligence wired, so transcripts become CRM deals with real values.

Delivered: Both channels running, and attribution you can put in a board deck.

Days 76–90

Handover

Your repo, your infrastructure, your keys. Team training. A documented playbook that survives the people who wrote it.

Delivered: The 90-day report, and the keys.

What you have at the end

  • Every meeting attributed, with the evidence behind it
  • Campaigns running on both channels, on a list where every company was verified
  • A cost per verified meeting you can calculate and defend
  • The whole system running in your own infrastructure, with your team trained on it

What I will not promise you

A meeting volume. Anyone quoting you one before they have seen your list, your offer and your domain reputation is guessing, and you find out which in month three. What I will commit to is that every meeting the system produces will be attributed with evidence. That is the number that settles the argument.

An agency costs $168,000 over two years. And you own nothing.

At $7,000 a month that is the bill, and when it ends they keep the playbook, the list logic and the copy library.

This is $25,000 once, then a few hundred a month in tools you pay the vendors directly. By month four you are ahead. By month twenty-four you are roughly $140,000 ahead, with an asset on your side of the table.

Book a call

We work out which of the three fits, or whether none of them do.

Who this is for

I take two build engagements at a time, so it is worth being blunt about fit.

  • A founder or revenue leader who wants an outbound engine they own, not a subscription they renew
  • A team already spending $5k a month or more on an agency or an AI SDR tool and unable to prove what it produced
  • Someone who tried to make outbound a rep’s side project and watched it stall
  • An agency founder who wants to productise this motion into a repeatable service

It is probably not a fit if you want a done-for-you outbound agency, and I will happily point you to good ones. Same if you are pre-revenue and still finding product-market fit, or if nobody internally can hand over five API keys inside a fortnight. Credential handover, not engineering, is what actually delays these builds.

The questions I get asked

Why not just hire a GTM engineer?
If you can find a good one and afford them, do it. The problem is that the title got hot faster than the skill spread, so most of what is available is someone who has watched a lot of Clay tutorials. You will not be able to tell the difference at interview, and you find out four months and a botched system later. Buying the built thing removes that bet.
We already use an AI SDR tool.
Then you already know what generic copy looks like, and you probably have no way to verify what it produced. Worth asking who actually replied to it. In my experience the answer is the accounts you least want, because they are the ones with nothing better to do than answer generic email.
Can’t we build this ourselves?
Yes, in about six months. I would point you at three specific mistakes that will cost you most of that: an ICP definition that rejects a company that had just become a customer, a CRM default that silently fabricates pipeline across 18 deals, and a persona targeting job titles that barely exist at the company size you are selling to. That last one cost 82 times the addressable market. I have made all three. That is most of what you are buying.
What is the difference between the MVP and the full build?
Scope, not quality. The MVP is one ICP, one persona, one offer angle and one email campaign, live in two to three weeks. It is the real system running on your business at small scale, and it is yours to keep either way. The full build extends it across your whole ICP set, adds LinkedIn, wires attribution and call intelligence through, and hands the whole thing to your team with training and a playbook.
What if we want you to keep running it?
Then I do, at $3,500 a month. It stays yours either way. That is the point of the model: you can end the arrangement and the machine keeps running.
How long until we see something?
The workshop is a day. The MVP has a campaign sending inside three weeks. On the full build your first campaigns are live around week six, and both channels are running by week eleven.
Does our stack change the timeline?
Yes. Attio, Sendkit and Calendly is the fast path. Instantly or Smartlead adds about a week. HubSpot pushes the first campaign to around week ten, Salesforce to around week twelve. The delay is credential handover and CRM schema archaeology, not engineering.
What breaks when a vendor changes their API six months from now?
It breaks in your repo, because you own it. That is the honest trade for ownership. Either take the maintenance licence at $1,000 a month and it is mine to fix, or your team owns it and I will make sure they are trained to. I will not leave that ambiguous in the contract.
Who owns deliverability?
I supply the mailbox infrastructure and build the system, so I take it seriously: warmup periods, volume ramps, and a written policy that nothing sends to a list that has not been QC’d. That policy is not negotiable, including when you are impatient.
What if the ICP does not work first time?
It will not. It did not on my own business, twice. Two calibration rounds are budgeted into the build phase and written into the scope, rather than billed to you as a surprise.

One day, $1,000, and you will know if this is worth doing.

The workshop is the cheapest way to find out whether a build makes sense for you. If it does, the fee comes off it. If it does not, I will tell you that on the day and you will still leave with your ICP written down properly.

Book the workshop

We work out which of the three fits, or whether none of them do.