ADSAgentic Data Service
Grounded agentic intelligence for the enterprise.
Autonomous agents can reason, delegate and act. ADS gives them the knowledge, boundaries and accountability required to do real work.
01The Platform
The system beneath autonomous work.
Any process can call a model. A system is the part that knows what an agent was allowed to touch, what it used, what it decided and what it changed. ADS is building that layer.
- S01
Governed knowledge
Structured domains over a typed ontology. Knowledge is curated, versioned, and owned by name.
- S02
Agent access
Agents reach the system through MCP, under the same permissions and scopes as any other consumer.
- S03
Memory
Durable working context that outlives the session and stays attached to the work.
- S04
Tools
Declared capabilities with declared effects. What an agent can do is written down before it does it.
- S05
Workflows
The unit of value. Repeatable work with an owner, a definition, a state and an outcome.
- S06
Validation
Shape and constraint checks on what enters the knowledge core and on what agents produce.
- S07
Provenance
Every assertion carries where it came from. Every action carries what it was based on.
- S08
Audit
Execution history that outlives the run, the model version and the person who started it.
- S09
Model routing
Models are interchangeable components. The record of the work outlives all of them.
02Knowledge Core
Knowledge before generation.
A model remembers what it was trained on. Your contract terms, your policy revision and the decision your team made last quarter were absent from that training set. ADS grounds autonomous work in structured knowledge that is validated, attributed and reviewed, so an answer can be traced back to the thing it came from.
- Domains
- Knowledge is scoped into owned domains. Each one has a named owner, and that ownership is enforced at the boundary.
- Ontology
- Entities and relationships are typed. An agent follows a named edge to reach a fact, and can say afterwards which edge it followed.
- Provenance
- Each assertion retains its source, its author and its revision. Everything that enters the core carries a name.
- SHACL validation
- Shape constraints are enforced on write. Malformed or contradictory structure is refused at the boundary and returned to its author.
- MCP tools
- Agents query and act through Model Context Protocol interfaces, under the permissions of the caller.
03Agents
Intelligence is no longer singular.
Agents specialise, delegate and hand work to each other. A route is refused. Two reroute. One completes. ADS is the governed environment they operate inside, and the record of what they did while they were there.
Specialisation is only useful if the specialists can reach each other. Delegation, handoff and escalation happen inside one system, under one set of rules. Autonomy without isolation.
Research
Grounds the work in validated knowledge, and says which sources it used.
Planning
Turns an objective into a governed path of steps with owners and limits.
Execution
Carries out the declared actions the permissions actually allow.
Validation
Checks what comes back (shape, constraints and policy) before it counts.
Memory
Keeps working context attached to the work, beyond any single session.
Guardrail
Watches the boundary: budget, permission and escalation, enforced below the agent.
04Workflows
Every workflow has an owner.
ADS is built around measurable work: a named owner, a governed execution path, and an outcome someone can evaluate afterwards. A capability is a feature. A workflow is a responsibility.
Insurance claims
OperatingFinance & leasing
OperatingWorkflow bundles
ADS is building workflow bundles: the knowledge, rules, views, agents and actions one piece of repeatable work needs, packaged so it can be deployed, governed and measured as a single object. Each bundle is built with the organisation that will own it.
- Knowledge
- Rules
- Views
- Agents
- Actions
- Evaluation
05Trust
Trust is a check that runs first. The rest is recorded.
Agents need room to act. Enterprises need to know what happened, why it happened, and whether it was allowed. A boundary enforced below the agent widens the range it can be trusted with: autonomy, without chaos.
Provenance
Every action resolves to the knowledge and sources it was based on.
Validation
Shape and constraint checks run before anything enters or leaves the core.
Permission
Scopes and permissions are enforced at the boundary, below the agent.
Policy
The rules of the workflow are written down and applied to every run.
Budget
Spend and usage ceilings per workflow, measured while the work runs.
Audit
Execution history that outlives the run, the model and the operator.
06Provenance
Every action has a history.
Run the execution backwards. The outcome resolves to an action, the action to a decision, the decision to the agent that made it, the agent to the memory and knowledge it stood on, and that knowledge to a source with a name on it.
- 01Outcomerecommendation issued
- 02Actionterms.prepare()
- 03Decisionwithin exposure limit
- 04Agentexecution/03 · guardrail/01
- 05Memoryportfolio.state@r118
- 06Knowledgeagreement.terms · validated
- 07Sourcemaster agreement · rev 7 · owner: Credit
The record outlives them.
Models are replaced. Prompts are rewritten. Teams change. The lineage of a decision survives all three, because it lives outside the model.
Worked example: trace for RUN-4417, Finance & leasing.
07System of Record
The system of record for agent work.
An operational surface where the work, the agents running it, the evidence behind it and what it cost are all the same object.
Active executions4 running
| Run | Workflow | Agent | Step | State |
|---|---|---|---|---|
| RUN-4417 | Finance & leasing | execution/03 | terms.prepare() | Running |
| RUN-4402 | Insurance claims | validation/01 | coverage.check() | Held |
| RUN-4398 | Leadership coaching | research/02 | session.prepare() | Complete |
| RUN-4389 | Insurance claims | guardrail/01 | evidence.request() | Complete |
ValidationShape constraints
Accepted this cycle1,284
Rejected on write97
Held for review23
ProvenanceRUN-4417
- Outcome · recommendation issued
- Action · terms.prepare()
- Decision · within exposure limit
- Agent · execution/03 · guardrail/01
- Memory · portfolio.state@r118
- Knowledge · agreement.terms · validated
- Source · master agreement · rev 7 · owner: Credit
Routing & budgetCycle
Token budget consumed61%
Routed · reasoning tier38%
Routed · fast tier62%
Illustrative interface · representative values · sample data
08Measurement
Measure what agents actually do.
A benchmark measures a model. An operations lead needs to know whether a workflow completed, whether the completion was allowed, and what it cost. ADS is building the instrumentation for those three: validation results and execution telemetry treated as first-class evidence.
Governed completion
Validation by task class
Agent failure modes
The evidence loop
- Execution
- Validation
- Rejection
- Learning
- Better routing
- Better execution
09Company
Building the operating layer for autonomous work.
AI is moving from answering questions to doing work. ADS exists so that the organisations making that move keep the knowledge it runs on, the rules it runs under, and the record of what it did.
Ownership
Autonomous workflows should remain owned by the enterprise operating them, including the knowledge they stand on and the record of what they did.
Grounding
Agents should act from governed knowledge. Where the ground is missing, the system should say so.
Accountability
The more an agent is trusted to do, the more precisely its work has to be reconstructable.
Simple rules. Remarkable systems.
10Request access
Put intelligence to work.
Tell us which workflow you want autonomous systems to operate.
We work with a small number of organisations at a time, building the knowledge, governance and execution path for one real piece of work before widening scope.