An agentic company.
One Flow at a time.
A Flow is a way of getting work done—like handling a support request, checking a payment or investigating an issue. It brings people, AI and software together, with your company's rules and approvals built in.
Built around real work in finance, support and engineering.
One useful company process.
Knowledge and controls you can reuse.
More of your company, Flow by Flow.
Build the company knowledge
your AI will need.
Useful AI needs context: your products, customers, decisions and the way your company works. Making that knowledge available is a foundation you can start building now.
TAC-OS helps you build a lasting corpus, control who can use it, and account for the work done with it.
A practical first step.
Room for the whole company.
A scheduled check. A support intake. A financial calculation. An issue ready for investigation. Choose one process worth improving and define the Flow around it.
Keep your existing systems. Give large language models the information and tools the task allows, with people at the decisions you choose. Every Run retains evidence and outcomes that later work can use.
Then add the next useful Flow, building on the knowledge, permissions and tools already in place.
Our approach to adoptionTurn a ticket into useful context
A customer reports an unexpected result
- Input
- The observation and its sharing policy
- Outcome
- Private evidence and selected search content
Keep the knowledge.
Improve the brain.
Flows leave source material, evidence, decisions and results. tacFS files and Run records build a lasting company corpus.
The brain is a working view built from that knowledge. People and agents can use it to answer questions and inform the next piece of work, which can add further knowledge.
TAC-OS is model agnostic. As models and methods improve, new Flows can build better brains from the retained evidence. Each new assessment keeps its own record.
A lasting foundation. Better ways to use it.
Put AI to work.
Know what happened.
Permissions define what AI can access and do. Human approval Steps record the decisions people make. The audit trail lets you account for the work that passes through TAC-OS.
- What is the AI doing?
- Follow the Run and its Steps: the plan, selected model, declared tools, reported progress and where work is waiting.
- What did it find?
- Inspect the returned assessment, source evidence and tool outcomes. Follow a conclusion back to the recorded material that supports it.
- Who accessed the information?
- Trace recorded access through TAC-OS to the user or agent identity, the information requested and the authority used.
- Was a human involved in the decision?
- Inspect the explicit human approval: who submitted the decision, what they approved and the evidence attached to that Step.
A safety net for company work.
Protected information. Explicit authority. Recorded human decisions. And retained results to support investigation and recovery when work needs attention.
Ask the company brain.
Through tacApp, people will be able to connect questions to company evidence and processes. Explore the answer, understand what supports it, then request the next useful Flow.
Choose a question
Which customer problems should shape our next release — and why?
Prioritise the offline sign-in investigation. It is the largest recurring group in the recent support cases, and there is already a candidate change to evaluate.
- Offline sign-in
- 8 related cases. An existing candidate fix is awaiting review.
- Interrupted downloads
- 5 cases. A fix has shipped; check which reports concern the newer version.
- Export options
- 3 requests. These describe additional functionality rather than a failure.
Follow the evidence
16 distinct cases grouped by recorded classification and source evidence: 8 offline sign-in, 5 interrupted downloads and 3 export requests. Repeated messages in a case are counted once.
Two crash signatures appear in reports attached to the sign-in cases. They are useful investigation inputs; the common cause has not yet been established.
The sign-in issue has a candidate change awaiting review. The download issue is linked to a released fix, so the next check is whether affected customers are using that version.
Investigate recurring sign-in failures
Use the eight cases, related crash evidence and candidate change to assess the cause and recommend the next engineering task.
- Compare the permitted case evidence
- Evaluate the candidate against the reported symptoms
- Return findings and a proposed task for human review
The proposal is reviewed before any code change or customer response.
Build towards
an agentic company.
Company knowledge, controlled execution and ways for people to take part. Explore the capabilities that give each useful Flow somewhere to grow.
Connected workflows
Bring agents, software and people into a process you can understand and operate.
ExploreCompany brain
Keep the source evidence and recorded work. Build better company brains as models and methods improve.
ExploreControl, audit & recovery
Know what AI did, what it found, who accessed the evidence and where people made decisions.
ExploreBench
Define, inspect, test and improve the way your company works.
ExploreOperations
Follow Runs, investigate outcomes and bring the right decisions to people.
ExploreModels & costs
Choose models around the work, keep the company corpus and adopt better methods as they emerge.
ExploretacApp
Ask the company brain, follow work and make the decisions that move it forward.
ExploreIntegrations
Bring issues, code, payments and customer evidence into connected company knowledge.
ExploretacFS
Keep source files and company documents with their versions, evidence and access rules.
ExploreYour systems.
A connected company brain.
Issues, code, payments, crashes and conversations all add context. Connect the sources each Flow needs and let the evidence become useful across more of the company.
Explore integrationsAlongside your own support system and company APIs.
Need an integration? Talk to us
Taking shape
at Anquet.
We're developing TAC-OS around the needs of a working software company. Financial processes, support evidence and code changes are helping shape the same underlying system.
Follow the workFollow the calculation.
Coordinate processing and retain the outcome.
Keep useful context.
Bring tickets and their evidence into the company corpus.
Move work through review.
Connect an issue, a candidate change and its checks.
What would you put to work?
We're interested in the processes you want to improve.