Coco
Register for Beta
Use cases//use-cases/find-design-partners
Use case · GTM motion

Find Design Partners with AI

Five ranked accounts in your Gmail drafts, with the research and sources attached. About 25 cr for the shortlist, 4–6 cr per draft — roughly 50 cr end to end, approval-gated. Nothing leaves your account until you click Approve.

Shortlist~25 crPer draft4–6 crFirst run3–7 minApproval-gated
Try Coco for free See the play
1,000 cr free· no card· ~2-min setup
Direct answer
~6 min read · 1,300 words

Coco builds your design-partner shortlist end to end: researches candidate accounts against your ICP, scores fit against problem and stage signals, and queues a personalized first-touch draft for each.

About 25 cr to shortlist 5–10 candidates from a target segment, plus 4–6 cr per draft. Every step you can review, edit, or send back. The result is a list of accounts you'd actually want as design partners — not a name-and-email dump out of an enrichment tool.

01

Design-partner search is its own job — not a filter run.

It's narrow. You want 10 to 30 accounts, not a million. Enrichment platforms are built for the opposite — filling pipelines, not finding a small set of accounts that match a specific operational pain.

It's high-context. The fit signal isn't "Series A fintech in NYC." It's "Series A fintech in NYC that just hired a head of data and mentioned their customer-data pipeline on a podcast last month." That's a research problem, not a filter problem. Apollo and ZoomInfo give you the firmographic match; they won't tell you which of those companies has the operational problem you actually solve.

It's also a one-time job in a busy week. Most founders build the design-partner list on a Sunday, get through eight accounts with real research, then send templated emails to the other twelve because Monday arrives. The good list compresses into the time-starved list, and the response rate reflects it.

Coco is built for the gap: not by giving you more names but by doing the research the names need to be worth using.

How Coco runs this play

Five steps in one chat. Approve at each beat.

You describe the ICP in plain English. Coco proposes the plan and the credit cost. You approve. Coco runs each step and surfaces the artifacts for review — every external action gates on you the first time through.

01

Source candidates.

Coco searches your connected enrichment tools for accounts matching the ICP. Firmographics first, then signal layers.

You might say

Find Series A fintech founders in the US, 50–200 employees, with a customer-data integration problem. Prioritize companies that have hired a head of data or platform engineering in the last 60 days.

Apollo LinkedIn Clay
Cost~8 cr
Time~1 min
Approve
02

Run a depth pass per candidate.

For each account that clears the firmographic filter, Coco does the kind of account research a careful human would: recent funding, hiring signals, public commentary on LinkedIn or in podcasts, news mentions, decision-maker backgrounds.

LinkedIn Web Podcast index
Cost~10 cr
Time~2 min
Approve
03

Score fit against your criteria.

Each candidate gets a fit score and a two- to three-sentence rationale. You see the exact signal that drove the score, not an opaque number. Disagree and Coco re-weights.

Coco · internal
Cost~5 cr
Time~1 min
Approve
04

Surface the shortlist.

Candidates ranked, with rationale and sources linked. Edit, drop with a reason, or send back for more. The next time you run this play, dropped patterns are remembered.

Coco · workspace
Cost~2 cr
Time~30 s
Approve
05

Draft outreach on approval.

Approve the shortlist; Coco drafts a personalized first-touch per account, anchored in the research, in your voice. Drafts queue in your Gmail Drafts folder. Send goes through your sender identity; nothing posts until you approve each send.

Gmail · drafts
Cost~5 cr / draft
Time~1 min ea.
Approve each
First run · shortlist plus five drafts.
~50 cr total8–10 min5 approval gates
What you get back

The shortlist, in your workspace.

A real run on the ICP above — Series A fintech with a customer-data integration problem. Five ranked accounts, fit scored, signals cited, sources linked. The same surface Coco hands back to you in-app, sized for the page.

Shortlist · Series A fintech / CDI problem5 of 5
Generated 22 May 2026 · 25 cr · 4 min 12 s
Awaiting your review
All 5Top fitDropped
Sorted by fit · desc
01Rank

Halberd Finance

halberd.com·Series A·$14M·80 employees·New York
  • Hired Marcus Avila as Head of Data on April 14, 2026— linkedin.com/in/m-avila
  • CEO Nina Levin on Founder Story podcast, Apr 28: "Our CDI pipeline is the bottleneck on everything"— ep. 144 · 32:10
  • Eight LinkedIn mentions of "customer data" in the last 60 days from the engineering org— search.linkedin
3 sourcesView →DropKeep
94/ 100 fit
02Rank

Telleron

telleron.io·Series A·$11M·120 employees·San Francisco
  • Hired Priya Shah as VP Platform Engineering on March 30, 2026— linkedin.com/in/priyashah
  • Public job listing for "Customer Data Integration Lead" posted May 2— telleron.io/careers
  • Customer story on their blog mentions data-pipeline rebuild as Q3 priority— blog/data-q3
3 sourcesView →DropKeep
91/ 100 fit
03Rank

Northpath

northpath.co·Series A·$9M·95 employees·Brooklyn
  • CTO Lior Bash on LinkedIn, May 9: "Our integration debt is now the top quarterly risk"— post · 412 reactions
  • Hired three senior data engineers in Q1— linkedin company page
  • Recent press: pricing pivot to enterprise plan likely raises integration complexity— fintechwire.com
3 sourcesView →DropKeep
88/ 100 fit
04Rank

Harbora

harbora.com·Series A·$7.5M·60 employees·Austin
  • Series A close announced January 2026 with a stated data-platform investment thesis— techcrunch.com
  • Engineering blog post on their CDP rebuild, published April 2— /blog/cdp-rebuild
  • Two public talks by CTO on customer-data plumbing in 2026— youtube + conf
3 sourcesView →DropKeep
84/ 100 fit
05Rank

Quarter Crossing

quartercrossing.com·Series A·$13M·140 employees·US-incorporated, London team
  • Director of Data Engineering hired Feb 19; bio explicitly mentions "fixing integration sprawl"— linkedin.com/in/d-okafor
  • Two LinkedIn mentions of integration cleanup from leadership in last 30 days— search.linkedin
  • Mentioned in Pragmatic Engineer as a fast-growing CDP buyer— newsletter · Apr 23
3 sourcesView →DropKeep
81/ 100 fit
5 accounts · 24 sources~25 cr spent4 min 12 s
Approve shortlist →

After you approve, drafts queue in your Gmail Drafts folder under your sender identity. Coco never sends without your second click the first time you run this play.

02

Fit scoring you can argue with.

Coco's fit scoring isn't opaque AI vibes. It's grounded in what you told it about the ICP, and the signals are inspectable.

If you said "prioritize companies that hired a head of data in the last 60 days," the score for that signal weights heavily, and any matching account has the hiring signal cited inline — "Hired David Park as Head of Data on April 7, 2026." You can disagree, but you can disagree about something specific.

If you said "prioritize companies with public commentary on their customer-data integration pain," Coco looks for podcast quotes, LinkedIn posts, or news mentions of the pain. Found ones get linked back to the source. Ones with no public commentary score lower on that dimension, and you choose whether to keep them in or drop them.

Coco surfaces what it found, with sources, and asks you to make the judgment. It doesn't tell you who your design partners should be. It does the research that lets you tell.

How autonomy unlocks

First run, versus once Coco's earned the work.

On day one, every external action gates on you. Once you've approved the same pattern a few times and the output earns the next job, you can authorize Coco to run the upstream steps autonomously — drafts still queue for review.

Every external action gates on your approval. Coco shows the artifact, you sign off, the next step runs.

You've authorized this pattern. Coco runs the upstream steps on its own and only pauses where you've told it to — before anything goes to a prospect.

  • 01Source candidatesApollo · LinkedIn · Clay Gated · approve Auto · within your rules
  • 02Run depth passPer-account research with sources Gated · approve Auto · within your rules
  • 03Score fitInspectable rationale per account Gated · approve Auto · within your rules
  • 04Surface shortlistRanked, sources linked Gated · approve Auto · within your rules
  • 05Draft outreachQueues in Gmail · your sender identity Gated · approve Still gated · always
The promise. Nothing leaves your account until you click Approve. Every send goes through your Gmail under your sender identity. Coco never spoofs anything. See how the four-step approval loop works for the trust model.
What stays gated, always. Sends never go autonomous, even on a trusted pattern. You can revoke the autonomy authorization in one click and Coco reverts to first-run posture. The autonomy is a budget you grant — not a setting Coco changes on its own.
Why not the obvious tool

What an Apollo list doesn't give you.

Apollo, ZoomInfo, and Clay all earn their keep. But none of them are doing the work design-partner search actually needs.

  • Apollo and ZoomInfo are enrichment platforms.They give you records that match firmographic filters — and that's a useful job. Coco uses Apollo as a lookup source in step 01. But a record isn't a design partner. The work is the research that turns a 200-row CSV into a 10-row shortlist with a defensible reason for each row. A record is the input to that work, not the output.
  • Clay is the DIY enrichment engine.Clay is powerful, and right for some teams. The trade-off is configuration time — you design the workflow, then run it. Coco's bet is the opposite: for design-partner search, where the workflow is roughly the same shape every time, the workflow ships pre-built. You tune the criteria; you don't build the criteria-evaluation engine.
  • The Sunday-afternoon manual pass.If you've got eight to ten accounts and three uninterrupted hours, your own research will out-rank anything an AI ships today. Coco's wedge is exactly the case where you don't have three hours — you have 45 minutes between meetings, and you'd rather come back to a shortlist than start with a CSV.
Common questions

Frequently asked.

01

How long does a design-partner search take?

Typically 3–7 minutes for a shortlist of 5–10 at standard research depth. The plan card shows the estimated time before you approve. Deeper research per account stretches the time and the credit cost; lighter research shortens both. Drafting adds about a minute per account on top.

02

Can I refine the ICP after the first run?

Yes. Tell Coco what's wrong with the shortlist — "drop anyone under 50 employees," "prioritize companies whose CEO is a repeat founder" — and Coco re-runs with the refined criteria. The constraint gets remembered for the next search in this workflow.

03

Will Coco send the outreach without my approval?

No. Drafts queue in your Gmail Drafts folder; sends gate on your approval the first time you run the pattern. You can authorize Coco to send for an approved pattern later, but the default for any new workflow is always approve-before-send.

04

What if the candidates aren't right?

Drop them with a reason — "not technical enough," "we don't sell to bootstrapped companies." Coco learns the constraint and applies it to subsequent searches in the same workflow. The shortlist gets sharper as you give feedback.

05

Can Coco do this for B2C design partners?

Coco is designed for B2B GTM. B2C design-partner search is possible in principle, but the fit signals Coco relies on — hiring signals, public operational commentary, decision-maker backgrounds — are sparser in B2C, so the output is less reliable. For B2C beta-user search specifically, Coco isn't the right tool today.

Get started

Hand Coco one ICP this week.

The shortlist takes under ten minutes, with the rationale per account you'd want before sending outreach. See whether it replaces the Sunday-afternoon pass.

Try Coco for free →Book a walkthrough →
1,000 cr free· no card· ~2-min setup