Orchestra: the AI agent team for travel and expense

Orchestra is the AI agent team for travel and expense: an AI project manager coordinating nine specialist agents that investigate, analyse and execute complete T&E projects from start to finish, returning finished, cited, editable work. It runs on the Cogent AI Framework, launched in June 2026, and every project passes five human-governed gates. You direct. The team executes.

Autonomous in execution, human-gated in decision. Once you approve the plan, the specialists run without supervision: they query your data, cross-check each other's findings, and hand results to the project manager, which synthesises the report. What they never do is decide alone. The plan waits for your approval before anything runs, surprising findings pause at a discovery checkpoint, and nothing commits spend without a named human. Autonomous AI agents are only useful in enterprise T&E if the autonomy is bounded, and Orchestra's is.

Workflow automation executes a path you designed in advance: if this, then that, every time. An AI agent team decides the path. Give Orchestra a goal and the project manager scopes it, drafts a plan specific to that question, dispatches whichever specialists it needs, and adapts when the data surprises it. Automation does a task. A copilot helps a person. Orchestra owns the outcome, and brings you the decision rather than the busywork.

Define, plan, execute, validate, deliver. You set the goal and Orchestra asks two to four scoping questions; a numbered plan waits for your edit and approval before anything runs; specialists execute in parallel in full view; surprising findings pause at a discovery checkpoint for your review; and a cited, editable deliverable lands in PowerPoint, Google Slides, Word, PDF or Markdown, re-running on your schedule.

Pre-built, curated templates ship at launch, with the catalogue being finalised with alpha clients; the air market analysis ships first. Each encodes a proven analytical workflow, so nobody has to become a prompt expert and every run follows the same structure. Any methodology can run on demand, on a weekly, monthly or quarterly cadence, or on event triggers when new data lands.

Three tiers run on every deliverable: silent validation detects and corrects errors before the draft, an independent audit checks data health, and a provenance appendix traces every claim to its source query. Surprising findings pause at a discovery checkpoint for your judgement before the report is written. The full validation architecture is documented on the Cogent AI Framework page.

For recurring analytical work, largely yes: market analyses, renewal benchmarks and QBR packs that cost $50K to $200K per engagement and arrive in weeks now run in minutes on your own data, on your cadence. Strategic advisory built on human relationships remains complementary. The practical shift is economic: depth and continuity replace periodic, expensive snapshots, and your team reviews finished work instead of commissioning it.

It changes what they build. In-house analytics teams spend most of their capacity servicing requests: the deck, the pull, the one-off question. Orchestra absorbs that queue, which frees your analysts for the modelling, data engineering and judgement work only they can do. On build versus buy, the model layer is not the hard part. The hard part is the consolidated T&E estate, the validation rail and the methodology library, and that is years of work specific to travel and expense. The honest comparison is usually not Orchestra versus a build, but Orchestra now versus a build in eighteen months.

Yes. Orchestra launched in June 2026 and is deploying across its launch customers. Because it runs on the Cogent AI Framework, organisations already on PredictX add it without new integration work: connectors and certifications carry over. Onboarding configures your workspaces, methodologies and schedules, and the first project runs on your own programme data during the walkthrough.