Meet the AI teammates inside Finta
Aurora Agents are AI teammates for recurring private-capital work. Give one a recognizable identity, a continuing responsibility, instructions for how you want the work done, and a place to report back. Aurora remains the primary Finta agent. You can add specialized teammates for work such as investor research, LP prospecting, meeting preparation, relationship follow-through, and diligence coordination.
The useful shift is not from one chat window to several chat windows. It is from asking for isolated outputs to assigning a job that has an owner.
An AI teammate should make five things easy to understand:
- Who it is: a name and character you can recognize.
- What it owns: one bounded responsibility.
- How it works: standing instructions, sources, and quality standards.
- What it brought back: a brief, shortlist, draft, status update, or other reviewable artifact.
- What needs you: a decision, permission, correction, or missing input.
That is the idea behind Aurora Agents. Meet the teammate, inspect the work, make the consequential decision, and return to the same working relationship later.
Aurora is the primary identity
Every Finta workspace starts with Aurora. She is the general agent for working across the relationship and deal context available in Finta. Additional agents sit underneath Aurora as optional owners of particular jobs.
You might create:
- a Researcher that prepares sourced manager or company briefs;
- a Prospector that maintains a focused list of potential investors or LPs;
- a Meeting Partner that prepares the context and open questions before an important conversation;
- a Follow-Up Partner that turns commitments and requests into drafts and visible next steps;
- a Deal Partner that keeps diligence questions, materials, owners, and review decisions organized.
These are not separate AI products. They use the same underlying Aurora system. The identity gives the work a stable owner, not a new permission boundary. Connected Apps, Skills, and workspace context remain governed at the Finta workspace level. A teammate should not be used as a substitute for separating clients, mandates, or information that requires distinct access controls.
For a deeper map of where these roles fit, read AI Teammates for Private Capital.
What changes when the work has an owner
A one-off AI request ends when the answer arrives. A responsibility continues.
Suppose a fund manager asks, "Which LPs should I contact?" A useful answer today is not enough to run the job next week. The work also needs criteria, exclusions, source dates, duplicate checks, relationship context, and a clear definition of what happens when evidence is missing.
Give that responsibility to a teammate and the operating question becomes more precise:
Maintain a focused list of LP prospects that fit our fund strategy. Explain why each institution belongs, preserve the public sources used, check Finta before adding a duplicate, and flag any prospect whose current mandate cannot be verified.
Now the teammate has a job. The shortlist is an artifact of that job. The fund manager still decides whether an LP fits, whether a relationship path is appropriate, and whether any outreach should occur.
OpenAI's current guide to agents describes an agent as a system that can manage a multi-step workflow using instructions and tools within guardrails. The teammate framing adds an operating layer that people can understand: a continuing role, recognizable ownership, and work that returns for review.
Four examples across private capital
The same teammate idea changes with the mandate. "Find investors" does not mean the same thing to a startup founder, a general partner, a family office, and a real estate sponsor.
| Person | Responsibility given to a teammate | Useful work returned | Decision that remains human |
|---|---|---|---|
| Startup founder | Maintain a shortlist of investors aligned with the round | Fit rationale, sources, existing relationship context, and next-step options | Which investors enter the active pipeline and how to approach them |
| Fund manager | Review the LP pipeline and prepare open follow-up | Updated evidence, unanswered questions, commitments, and draft follow-up | LP fit, allocation expectations, and external communication |
| Family-office principal | Prepare a brief before a manager meeting | Mandate summary, team changes, relevant materials, inconsistencies, and questions | Manager selection, diligence judgment, and investment decision |
| Real estate sponsor | Keep investor questions and approved materials connected | Question log, source material, responsible owner, and review-ready response | Disclosure, suitability, offering decisions, and sending the response |
The teammate owns follow-through on the assigned work. It does not own the relationship, fiduciary responsibility, investment judgment, or legal decision.
The first useful assignment
Start with a responsibility that is recurring, evidence-based, and easy to review. Do not begin by asking an agent to "handle fundraising." That scope contains too many different decisions, sources, and authority levels.
A better first assignment has six parts:
- Responsibility: the continuing job the teammate owns.
- Context: the records, documents, conversations, or public sources relevant to the job.
- Output: the artifact it should return.
- Quality: what a useful answer must include.
- Cadence: when or why the work should happen again.
- Review: which decisions or external actions require a person.
For example:
Before each LP meeting, prepare a one-page brief from the available CRM record, approved notes, relevant documents, and current public sources. Separate verified facts from inferences. Include the last interaction, open questions, and three discussion points. Ask me when a source conflicts or a requested document is not approved for sharing.
Use the copyable framework in How to Delegate Work to an AI Teammate to define your first job.
Identity helps people manage the work
Names and characters are useful because they make ownership legible. A teammate called Mira can be responsible for manager research. Another called Rowan can prepare LP meeting briefs. A user can find the right conversation, see which responsibility produced an output, and adjust the relevant instructions.
Personification does not make an agent competent. A strong teammate still needs:
- a bounded role;
- useful and permitted context;
- clear instructions;
- tools appropriate to the job;
- evidence in the returned work;
- visible status and blockers;
- a person accountable for consequential decisions.
Asana's current AI Teammates product explanation similarly organizes agents around roles, guidance, skills, access, and human review. That is useful evidence that "AI teammate" is becoming understandable product language. It is not an industry standard, and Finta's application is specific to the relationship-heavy work of raising and deploying private capital.
What Aurora Agents should not imply
The teammate metaphor becomes misleading when it hides the system's limits. Keep these boundaries visible:
- An avatar is not a permission boundary. Use workspace controls and operating policy for access decisions.
- A generated claim is not verified research. Preserve sources and recheck time-sensitive facts.
- A relationship path is not consent. A connector chooses whether to make an introduction.
- A draft is not a sent message. Confirm the status of external communication.
- A routine is not judgment. Scheduled work can prepare a decision, but it does not assume investment or fiduciary authority.
- Persistence is not constant activity. A teammate can resume from saved instructions, artifacts, and state without running continuously.
For the practical definition behind these boundaries, see What Is an AI Teammate?.
Give one recurring responsibility a teammate
Choose the recurring job that repeatedly makes you reconstruct the same context. Define the output you want back, the sources the teammate may use, and the point where human judgment takes over. Then improve the instructions from real work rather than trying to anticipate every exception on day one.
Explore Aurora to see how Finta carries available relationship context into a reviewable next step, then use the private-capital guide to choose the first teammate your workflow needs.
Research checked September 27, 2026. Product behavior, availability, connected sources, tools, and controls can change. The examples are illustrative workflows, not customer results, investment advice, or a promise that an agent will reach the same conclusion from incomplete data.