AI partner development · GTM

An AI partner manager that learns from who you pay.

Souk finds partners, runs their deals, and pays their commissions. I designed the loop that closes it. Every payout teaches the product which kind of partner to source next.

Role
Product design, end to end
Timeline
2025–26
In production at
Deel, Vanta, Google
Souk's sourcing screen: AI-matched partners with a fit score, a reason for each, and a personalized outreach draft
The money shot. Souk sourced eight partners for the account, scored each against how the team actually closes, and drafted the first outreach for the top match. The manager just reviews and sends.
Sourced for Beacon
Partners worth reaching out to
5
to contact
Ravenscroft AdvisoryConsultancyTop match
Sells into the same 2,000–5,000-seat buyers you target
94
fit
The RevOps CollectiveCommunity
12 warm intros into fintech ops leaders
91
fit
Marta DevlinCreator
Her list mirrors your closed-won profile
88
fit
Brightwater ConsultingAgency
Can bundle you into deals she already runs
85
fit
Harbor & ValeAgency
Overlaps with two of your open opportunities
82
fit
Ranked by how you actually close, not by reach.
5→1
tools replaced
100%
deals attributed to a partner
0
payout spreadsheets
1
loop that learns

Context

Souk is an AI partner development manager. It runs the whole partner lifecycle for GTM and RevOps teams: sourcing the right partners, tracking the deals they bring, and paying their commissions. It is in production at Deel, Vanta, and Google. I was the product designer, and I owned the three surfaces that make up the loop: Find, Pipeline, and Payouts.

The challenge

A partner program is one job that companies run across five tools. A CRM for accounts, an outbound tool for sequences, LinkedIn for sourcing, a contract app for agreements, and a spreadsheet for who gets paid.

Every handoff between them is a manual re-key, and nothing ties the pieces together. A deal closes in the CRM and no one can say for certain which partner brought it. Payouts get calculated by hand from a spreadsheet that is always slightly wrong. The whole program leaks at the seams.

In the field

I sat with partner managers and watched the day go. Most of it was admin. Copying a name from LinkedIn into the CRM. Chasing a contract. Reconciling a payout spreadsheet against deals that may or may not have been attributed right. The one question that actually mattered, which partners are worth more of my time, was the one thing none of the tools could answer, because the money they paid out never talked to the partners they went looking for.

The reframe

Stop building five tools. Build one loop where the money you pay out teaches you who to find next.

The design job was not five better tools. It was one loop with no seams. Find sources a partner, Manage runs the deals with attribution built in, Pay sends the commission from the revenue that partner closed. And then the loop closes: the payout data tells Find which partner types actually produce, so the next search is smarter than the last. One system that gets better every time it pays someone.

Before and after: five disconnected tools with leaky handoffs, next to one Souk loop where payouts feed back into sourcing
The shift, side by side. Before, five tools and four manual handoffs with no clean attribution. After, one loop where what you pay out decides who you source next.

How the loop holds

Find is only the opening. Two moves make the rest hold together: attribution that never leaks, and a payout that teaches the next search.

Move 01

From a leaky CRM to attribution that holds

Sourcing is only useful if the deals that follow trace back to the right partner. I built the pipeline around attribution: sourced, onboarding, first deal, producing, with the revenue owned by the partner who brought it. Nothing gets re-keyed between stages. The partner going cold surfaces on its own, so a manager works the row that needs them instead of scanning the whole list.

A partner pipeline from sourced to producing, each partner carrying its stage and the revenue attributed to it
Sourced to revenue in one view. Every deal is owned by the partner who brought it, and the one partner going cold is the only row that turns amber.

Move 02

From a payout spreadsheet to a loop that learns

The last step is where most tools stop: pay the partners. Souk takes each cut from attributed revenue and pays it automatically. Then it does the thing a spreadsheet never could. It reads who earned the most and feeds that back into sourcing, so the next Find leans toward the partner types that actually produce. Paying a partner is also how the product gets smarter.

A commission payout run, most partners paid automatically, with a note that payout data feeds back into partner sourcing
The loop closing. Commissions pay out on their own, and the earnings feed back into Find so the product sources more of what pays.

What changed

The loop replaced the stack. Partner teams at Deel, Vanta, and Google run sourcing, deals, and payouts in one place instead of five, and the manual re-keying between them is gone.

“A partner program was never five tools. It was one loop that no tool had bothered to close.”

Attribution is the part that made it hold. Because every deal is owned by the partner who brought it, payouts are exact instead of a monthly reconciliation, and the same data feeds sourcing. The product learns which partner types pay off and finds more of them. It gets better the longer a team uses it.

Reflection

Souk taught me that the product is the loop, not the screens. Anyone can design a sourcing tool, a pipeline, and a payout table. The design work was making the last step feed the first, so paying a partner is also how the product learns who to find next.

The senior move was refusing to ship three good tools when the job was one loop that closes on itself.

Back to work