What is agent trajectory data, and why do business records resemble it?
Agent trajectory data is a recorded sequence of steps, tool calls and outcomes for a task, used to train and evaluate AI agents. Business histories such as resolved tickets, deal records and approval chains often resemble trajectories because they log goals, actions and results as work happens.
What is agent trajectory data?
Agent trajectory data is a recorded sequence of steps an agent, or a person, took to complete a task: the goal, each action, the tools used, what came back and the final outcome. AI developers use trajectories to train and evaluate agents that carry out multi-step work. Company histories such as ticket threads and approval chains often look like trajectories because they are logged naturally as work happens.
What does a trajectory look like?
Illustrative example, fully fictional: a support ticket at a made-up software company.
| Step | Element | Example |
|---|---|---|
| 1 | Goal | Customer cannot export invoices |
| 2 | Action | Agent reads the ticket and account history |
| 3 | Tool call | Looks up the billing record in the finance system |
| 4 | Observation | Export is blocked by an expired permission |
| 5 | Action | Escalates to engineering with a summary |
| 6 | Tool call | Engineer resets the permission |
| 7 | Outcome | Customer confirms export works; ticket closed |
The sequence plus the outcome is what makes it useful. A model can learn which actions led to a resolved result and which led to rework.
How is it different from instruction tuning data?
| Data type | Shape | Typical use |
|---|---|---|
| Instruction data | One prompt and one answer | Teach models to follow requests |
| Agent trajectory | Many steps, tool calls and observations | Teach models to complete tasks |
| Expert trajectory | Steps by a skilled professional | Show high-quality task execution |
| Evaluation set | Task plus known correct result | Test whether an agent succeeds |
| Ground truth | Verified correct answer | Score outputs; see what ground truth data is |
Instruction data is short and one-turn. Trajectories capture the middle of the work, which is where agents fail.
Why do business records resemble trajectories?
Companies log work in the order it happens. A CRM tracks a deal from lead to closed-won. A ticketing system tracks a request through triage, escalation and resolution. An approval workflow records who decided what and why. Together with email and chat, these records show decisions, exceptions and handoffs, the same elements agents need to learn.
Negotiation threads are a clear case, covered in negotiation threads as AI agent training data. The explainer AI agents need work data lays out the demand.
What makes a trajectory valuable?
- A clear starting goal and a recorded outcome.
- Multiple connected steps across systems, not one isolated message.
- Tool use, such as lookups, updates and approvals, visible in the log.
- Quality: the work was done by competent staff and the outcome is known.
- Rights: the company created the records and can license them.
What does this mean for a referral partner?
You do not need to produce or describe any trajectory. Partners never export, upload or describe confidential records. Your role is to recognize a company with long, connected histories across 10-15+ systems and to introduce the sponsor. The company then builds an inventory with SourceX, with redaction rules agreed before any work begins.
Myths about what counts as valuable are tested in AI data licensing myths vs facts. To see how licensing deals are structured, read enterprise AI data licensing deals.
What should I say to a client?
What are the limits?
Trajectories drawn from consumer personal data, protected health information or client-owned work need consent or de-identification first. Records generated by AI only to be sold do not qualify. Whether a buyer wants a specific set depends on demand.
How are partners rewarded?
Partners earn 25% of the eligible platform fees SourceX actually collects from the referred company's licensing deals, capped at $100,000 per referred company. The reward is payable only after the buyer pays and SourceX receives its fee, and no reward is guaranteed. It is a share of SourceX's fee and is never deducted from what the company receives.
Next step
Check whether the company logs outcomes across systems with the company fit checker, and read how it works. Then register as a partner and introduce the owner or CFO.
- Step 1Share your linkSend your personal link to a company you know.
- Step 2Company appliesThe company applies itself at /apply.
- Step 3Buyer selects and paysThe buyer selects and pays for the data and SourceX receives its fee.
- Step 4You get your rewardYour share of SourceX fees becomes payable.
Common questions
What is the difference between agent trajectories and chat logs?
A chat log is a conversation. A trajectory records a task: goal, actions, tool calls, observations and outcome. Some chat logs contain trajectory elements, but a trajectory is defined by recorded steps and a known result rather than by the format of the text.
Do companies already have agent trajectory data?
Many have records that resemble it, such as resolved tickets, deal histories and approval chains. They were not created for AI training, so they usually need an inventory, rights review and redaction before a buyer can use them.
Is agent trajectory data the same as computer use data?
Computer use data is one kind of trajectory, capturing screen actions such as clicks and typing. Trajectories also include tool calls and API steps. Business system logs and ticket histories are a different, text-based source of multi-step task records.
Can any company's logs qualify?
No. Buyers look for connected, outcome-labeled histories, rights to license and enough scale. A company must also meet the baseline of 50+ full-time employees at peak (contractors excluded), several years of operations and an authorized sponsor.
Does a partner need to understand training to introduce a company?
No. A partner needs to recognize long, connected records and make the introduction. SourceX handles qualification, inventory, price and terms, buyer review, delivery and payment, and the company approves every step.
Related pages
Free resources
- IRR calculator — Internal rate of return on annual cash flows.
- Business valuation calculator — Enterprise and equity value from EBITDA, your multiple, cash and debt.
- Portfolio data opportunity scanner — Screen several companies in one session.
- All free tools · MCP resource center
By SourceX Partnerships Team · Published 2026-10-09 · Updated 2026-10-09
Know a US company with valuable proprietary data?
Become a referral partner from anywhere we support, get your link and introduce an owner or authorized decision-maker.
Refer a company →I own a business
Explore licensing your company's data to AI developers worldwide. Start a short assessment; no uploads needed.
Start an assessment