Your custom negotiation co‑pilot

A custom agent trained on your fee history, your past negotiations and the live 500+ firm benchmark — supporting every proposal and counter your team sends.

Start free: consult + benchmark report → How it's engineered →

From their counter to your reply — in five steps.

Paste the client's response; the co-pilot prices what it would cost, gives you the exact terms to send back, and drafts the reply — every number grounded in your own history and the benchmark. Click through the five screens — step 3b is an optional detail page.

1 / 5 Step 1 · Paste response
Click the screen to enlarge · click left / right to move through the flow
1The system

Eight parts. The AI is one small box.

1 The client’s counter lands The mandate, plus the pasted email: “our procurement team suggests 22%…” UNTRUSTED TEXT read as evidence to answer — never as instructions to follow 2 The full financial impact is calculated Every ask in their counter, priced: fee rate, salary base, retainer, guarantee — what accepting as-is would cost you. This is the “See impact” screen — page 2 of the flow above. 3 Your and Maksfi’s relevant data is pulled · Fees — your own closes, and the Maksfi fee database · Other T&Cs — salary bases, retainers, guarantees · Expected (full) compensation — down to secondary benefits Your data stays in your firm’s workspace — never pooled. THE MOAT your history plus our data — not the prompt 4 · The AI Agent one box of eight engineered prompt with subagent orchestration 5 A checkable answer — not loose prose Fee · basis · rate · sources · rationale — and the reply: the draft, the tactic behind it, and the suggested send window for the follow-up email. Prose can’t be audited; this can. what to say, what to hold back — and when to send 6 Double-checked — the maths, and your policy · Arithmetic recomputed independently — to the cent · Every claim must point to a source that actually exists · Minimum fee & rate, maximum discounts — your policy floors Fail either check, and the draft stops here. a failed check stops the draft — you never see it 7 Your consultant approves — or doesn’t Approve, edit, or decline. Nothing is priced, promised or sent without a person at your firm saying yes. there is no path from the AI to your client 8 The outcome is logged — and learned from Did the client accept? At what rate? Which reply — and which timing — worked? Every answer sharpens the next recommendation. The compounding bit Every outcome — fee and tactic — makes the next better. Your negotiation counterpart can copy box 4 in an hour. outcomes feed retrieval — the co-pilot compounds
1The client's counter lands

The mandate, plus the pasted email: "our procurement team suggests 22%…"

read as evidence to answer — never as instructions to follow
↓
2The full financial impact is calculated

Every ask in their counter, priced: fee rate, salary base, retainer, guarantee — what accepting as-is would cost you.

the "See impact" screen — page 2 of the flow above
↓
3Your and Maksfi's relevant data is pulled

Fees — your closes and the Maksfi fee database. Other T&Cs — salary bases, retainers, guarantees. Expected (full) compensation — down to secondary benefits. Held in your firm's workspace, never pooled.

the moat: your history plus our data — not the prompt
↓
4The AI Agent

One box of eight.

engineered prompt with subagent orchestration
↓
5A checkable answer — not loose prose

Fee · basis · rate · sources · rationale — and the reply: the draft, the tactic behind it, and the suggested send window for the follow-up email.

what to say, what to hold back — and when to send it
↓
6Double-checked — the maths, and your policy

Arithmetic recomputed to the cent, every claim source-checked, and your pricing policy enforced: minimum fee and rate, maximum discounts. Fail either check, and the draft stops here.

a failed check stops the draft — you never see it
↓
7Your consultant approves — or doesn't

Approve, edit, or decline. Nothing is priced, promised or sent without a person at your firm saying yes.

there is no path from the AI to your client
↓
8The outcome is logged — and learned from

Did the client accept? At what rate? Which reply — and which timing — worked? Every answer sharpens the next recommendation.

your negotiation counterpart can copy box 4 in an hour
↺ outcomes flow back into part 3 — the co-pilot compounds
2What that buys you

Four commitments, built in.

Your data stays yours — your fee history lives in your firm's workspace, never shown to another firm. part 3
No invented numbers — every figure and tactic traces to a source; "no comparable" beats a guess. parts 4–6
You stay in control — a consultant approves every send; no path from the AI to your client. part 7
Better negotiation every month — every outcome, fee and tactic, feeds back in. part 8
3Quality, twice over

Every draft faces a critic first

Inside the system, the draft that reaches your screen has already been through a review loop: a second, independent pass scores it against a fixed standard, and a draft that falls short goes back for revision — automatically, before you ever see it. The reviewer can reject; it can never rewrite. Author and judge stay separate, the same way you'd run it with people.

You only read drafts that have already survived scrutiny.

The standard every draft is scored against
  • Is every claim one we can evidence to the client?
  • Do the numbers match their sources, exactly?
  • Is the position defensible against your own track record?
  • Does it read like your firm — not like an AI?
  • Is the timing right — the response window and the follow-up?
  • Would a procurement lead take it seriously?
The back-test, step by step
  • Take a set of mandates your firm has already closed
  • Hide their outcomes from the co-pilot completely
  • Have it price each one from what you knew at the time
  • Compare its recommendations to what actually happened
  • You see the match — and the misses — before go-live

Tested on your history, before it touches your future

Anyone can demo an AI on a cherry-picked example. The honest test is a back-test: we run the co-pilot against mandates you have already closed — with the outcomes hidden — and show you how its recommendations compare to what your partners actually negotiated.

If the numbers don't hold up on your own history, you'll know before you've relied on it once. That's the standard we think AI in a fee conversation should have to meet.

4Trust practices

What's in place, what's in build.
Stated plainly.

In place today In build
  • Benchmark shows only anonymised aggregates — and only where at least 10 datapoints sit behind a figure
  • All traffic encrypted (HTTPS); all code under version control
  • Per-firm workspace isolation; no other firm's name can appear in your outputs
  • Arithmetic recomputed in code, every claim source-checked, abstention built in
  • Human approval before anything is client-ready

No SOC 2 or ISO 27001 certificate yet — we publish this status instead of implying otherwise. Ask us about any line.

5Fair questions

The questions we'd ask, answered.

Does our fee data help train something competing firms use?
No. Your mandate-level data stays in your firm's workspace and is never shown to another firm. The shared benchmark only ever displays anonymised aggregates — and only where a figure has at least 10 independent datapoints behind it (n ≥ 10), so no single firm's number can ever be read out of it.
What happens when the co-pilot doesn't know?
It says so. If there is no comparable placement in your history and no defensible benchmark band, the draft states that plainly and narrows to what can be evidenced. We consider a confident guess a defect, not a feature — abstaining is engineered in.
Can the AI contact our clients?
No. The co-pilot produces drafts and rationale for your team. A consultant approves — and usually edits — everything before it goes anywhere. There is no path from the co-pilot to your client that doesn't run through a person at your firm.
Which AI models do you use?
Commercial-grade models from established providers, run under a data-processing agreement, with the specific model disclosed to every client firm. We commit to providers whose terms bar training on your data. Final selection happens per pilot — and you'll know exactly what's running.
Are you SOC 2 or ISO certified?
Not yet, and we won't dress the site up to suggest otherwise. We're a young company building toward that bar. Until then our trust practices above are stated plainly — in place, in build, or committed — and we'll answer questions about any of them directly.
How do we know it actually works?
You don't take our word for it — you back-test it. Before your firm relies on the co-pilot, we run it against mandates you've already closed, outcomes hidden, and show you how its recommendations compare with what your partners actually negotiated. Go-live is earned on your own history.

Judge the system on your numbers.

Start where the co-pilot starts: the free benchmark consultation — 30 minutes plus a written report of your terms against the market. Yours to keep.

Start free: consult + benchmark report →