For companies, not individuals. An agency's most useful records are its creative loop: the brief, the first draft, the feedback, the revisions and the sign-off. Buyers say they want this kind of decision history. The limit is ownership, because the finished work usually belongs to the client.
Last checked: 7 October 2026. Buyer terms are quoted as published on the buyers' own pages.
Every project leaves a trail of what was asked for, what was tried, what was rejected and why.
Goal, audience, constraints, deadline.
Routes explored, some dropped early.
Creative director and client comments.
What changed in response, round by round.
Sign-off, and the reason it was accepted.
Finished creative work is everywhere. The written record of why a draft was not good enough yet is much rarer. An agency's project history captures exactly that judgment: a creative director's note that a headline is too clever for the audience, a revision that fixes it, and a client approval that confirms the call. micro1's data partnership page lists project histories, QA processes, "decision-making patterns" and "AI performance feedback", meaning human feedback on AI outputs, among what it wants. If your team already reviews AI-generated drafts and writes down what is wrong with them, those comments are that last category.
The tools agencies use appear on the buyers' source lists: Figma for design, Google Drive and Dropbox for files, Slack for chat, Asana and Monday.com for project tracking, HubSpot for the agency's own sales pipeline, Zoom recordings and transcripts for reviews, and Gmail or Outlook for client correspondence. Project management histories are especially useful because they timestamp each step of the loop.
In practice the loop is spread across systems, and each one holds a different slice. An Asana or Monday.com task records the brief, the due dates and the handoffs between account, creative and production. Figma keeps comment threads pinned to specific frames and a version history showing what changed after each round. A Slack review channel holds the internal debate before anything goes to the client. Drive folders hold the decks, and Zoom recordings hold the review meetings themselves.
Not every slice is equally usable. Figma comments are valuable but sit on top of artwork the client may own, so the text has to be separated from the frames. Client-shared Drive folders are the client's space as much as yours. Recorded review calls carry client voices and faces, and recording-consent rules differ by state. The internal slices, meaning task histories, internal review channels and your own playbooks, are the ones an agency is most likely to control outright.
Many agency contracts assign finished work to the client, and some assign all work product, including rejected concepts. Your contracts decide each row, not this table.
| Material | Often controlled by | Typical status for a sale |
|---|---|---|
| Final deliverables: ads, sites, videos, copy | The client, under assignment or work-for-hire terms | Out |
| Client brand assets, product details, launch plans | The client, often under NDA | Out |
| Rejected concepts and early drafts | Depends on the contract | Check each contract |
| Internal review comments and revision notes | The agency, but they describe client work | In, after de-identification |
| Stock images, fonts, licensed music | Third-party licensors | Out |
| Footage and photos of people | Shared: client, talent releases, the agency | Out unless releases allow |
| Brief templates, playbooks, SOPs, estimating guides | The agency | In |
The pattern is that an agency can usually license what describes its process and rarely license what it delivered. That is not as limiting as it sounds. Buyers say they want how work gets done, and the process side of an agency is where that lives. The harder question is the middle rows: internal comments and revision notes are written by your staff but are about client material, and the drafts they refer to may belong to the client. Client confidentiality obligations work the same way here as in law and accounting; our page on client confidentiality and data sales covers the shared rules.
Footage deserves its own caution. A shoot can involve the client's ownership, a production company's terms, talent releases and music licenses all at once, and a release signed for an ad campaign may say nothing about AI training. Our guide to video footage and images covers rights in people shown on camera.
A practical way to sort an archive is client by client rather than system by system. Pull the master services agreement for each client, note whether it assigns final deliverables or all work product, and note any clause on confidentiality or reuse. Clients whose contracts say no are excluded from every system at once: their Asana projects, their Slack threads, their Drive folders. What remains can then be scoped by system, with a much shorter list of judgment calls.
Agencies have no industry-specific line. The general published rules apply.
| Program | Published company payout | Published eligibility |
|---|---|---|
| Mode | "$100K-$5M" | 20+ full-time US office employees; several years of records the company owns; US-based teams the strongest fit |
| micro1 | "$100K-$2M+ for approved data packages" | 30+ employees (referral posting: 30 to 200); mature operations, documented processes, modern software tools; primarily English; US prioritized |
| Grepped | "$20K-$5M" | Any vertical; also pays individual professionals for expertise |
Last checked: 7 October 2026. Sources: data.mode.inc, micro1.ai/data-partnerships and micro1.ai/company-referral, grepped.ai, each as published on that date. Ranges are published across all sellers and are not offers.
Two details matter for agencies. Mode's rule counts full-time US office employees, so a studio that relies on freelancers may have fewer qualifying staff than people on its projects. And Mode asks for records "the company owns", which is the ownership question above in one phrase. micro1's "documented processes" fits agencies that run on written playbooks and tracked projects. The eligibility checker applies these rules to your numbers in your browser.
Agencies are lightly regulated as a sector, so the legal questions come mostly from intellectual property and contracts. Raise these by name:
Questions for your lawyer: Which client contracts assign all work product rather than final deliverables only? Do our NDAs cover internal comments about client work? Do any releases mention AI or machine learning use? What rights warranty can we give a buyer?
General information, not legal advice. Talk to your own lawyer before you sign.
The agency, its clients and its decisions are invented to show the method.
Example Creative Co. is a made-up agency with 34 full-time staff in a US office. It runs projects in Asana, designs in Figma, stores files in Google Drive, chats in Slack and tracks new business in HubSpot.
Published rules against its facts. Mode's 20+ full-time US office line is met. micro1's 30+ line is met, inside its posting's 30 to 200 band. Grepped lists any vertical. None of this is acceptance.
| Item | Decision | Reason |
|---|---|---|
| Process playbook, brief templates, estimating guides | In | Agency-owned, no client content. |
| Asana project histories | In, partly | Client names replaced; projects for four clients whose contracts forbid any reuse left out. |
| Slack creative-review channel | In, after review | Text comments kept; attached client images and files removed. |
| HubSpot new-business notes | In, reasoning only | Why deals were won or lost kept; contact details removed. |
| Figma files and final deliverables | Out | Assigned to clients. |
| Rejected concepts | Out | Most contracts assign all work product. |
| Video shoots with talent | Out | Releases do not cover this use. |
What is left is the agency's method and its judgment, recorded over six years. It would share a manifest and samples, approach more than one buyer, and compare terms as well as price.
Notice what drove most of the decisions: not the buyer's rules and not the systems, but the client contracts. The agency spent most of its scoping time reading agreements, not exporting data. That order is right. A narrower package built on material the agency clearly controls is easier to warrant, easier to de-identify and easier to explain to clients than a larger one with open questions in it.
All three are invented. They show how size changes eligibility, the shape of the archive and the number of contracts to read. Illustrative, not an offer.
A studio this size often runs on a handful of long-term clients, so one restrictive contract can remove a large share of the archive. Process tends to live in the founders' heads rather than in written playbooks.
Grepped's note that it pays individual professionals for expertise may matter more here than a company sale: a founder's judgment is the studio's main asset.
Enough clients that a few exclusions leave a usable archive, and usually a project management tool that has been used consistently for years. Account, creative and production teams each leave their own trail in the loop.
The main work is the contract-by-contract sort and stripping attachments out of comment threads.
Several brands or offices raise a basic question first: which legal entity holds which records, and which entity signed each client contract. The data license has to be signed by whoever actually controls the material.
Large brand clients are also more likely to have procurement terms that restrict any third-party use of their information.
Clients of an agency have communications teams of their own. Assume they will notice.
A brand that learns its agency licensed data to an AI company will ask one question first: was our work in it? If the answer is no, say so before they have to ask. If the answer is "parts of the process around it, with your name removed", the honest course is to tell them before signing, and for any client whose contract restricts use of its information, to ask rather than tell. Some contracts also require notice before client information goes to any third party, which a data buyer would be. Keep a one-page summary of what was in scope, per client, so account leads can answer consistently.
Generic questions for every seller, with the ones that bite agencies first.
Before any of this, build an exclusions list and inventory; prepare your data for sale walks through it.
Independent site. Some links are referral links: if your company signs with a buyer through them, the buyer may pay us a fee. You are not charged, and we never see your data.
We are not a partner, agent or representative of any buyer, and we cannot promise acceptance, amounts or timing. Practitioners cite 60 to 90 days to close. Compare terms on buyer programs compared first.Usually not. Many agency contracts assign finished deliverables, and sometimes all work product, to the client. Check each master services agreement and statement of work with your lawyer before treating any client work as yours to license.
The loop of brief, draft, feedback, revision and approval, plus the agency's own playbooks and SOPs. micro1 lists project histories, QA processes, decision-making patterns and AI performance feedback among what it wants.
Mode lists 20+ full-time US office employees. micro1 lists 30+ employees, 30 to 200 in its referral posting. Grepped lists any vertical with no size line. As published, checked 7 October 2026.
Treat them as out unless the client contract, model releases and any stock or music licenses clearly allow it. Rights of people shown on camera are a separate question from who owns the footage.
Nobody can say without reviewing the data. Buyers publish company ranges, not agency prices: Mode $100K-$5M, micro1 $100K-$2M+ for approved data packages, Grepped $20K-$5M. These are not offers.
Some contracts require notice before client information goes to any third party, and some restrict reuse entirely; your lawyer can say which of yours do. Even where nothing requires it, telling clients before they hear it elsewhere is usually the safer course for the relationship.
Mode's published rule says 20+ full-time US office employees, and micro1 publishes 30+ employees. Ask during application how either treats freelancers and contractors, and check the current wording. As published, checked 7 October 2026.