The product
An agent is more than a model.
There is no software to buy. Every agent we deploy comes with the same six components behind it, built for your process and delivered inside your environment: integration with your systems, governance that authorizes and audits every decision, interfaces to review and override its work, client-owned security, and the technology running it all.
It is deployed in a dedicated subscription inside your Azure environment, where your data and processing stay by default, and everything is licensed to you permanently. This is what turns an AI pilot into an enterprise-ready process.
Six components
The six components of every agent we deploy.
Agents
The three categories (Operations · Customer-Facing · Knowledge), the execution architecture, and how each agent owns one specific job in your business.
See how agents work →Integration
Six architectural patterns, six engineering principles, and the security model behind how we connect to your systems. No connectors: an engineering discipline.
See how we connect →Governance
The Governance Envelope. HITL by architecture on every deployment. Traceable test cases, scenario and regression testing before every release. Decision provenance for every action.
See the Governance Envelope →Interfaces
The operator surface your team uses day-to-day (dashboards, alerting, HITL controls, runbooks) designed per deployment. How your people stay in control of autonomous operations.
See the operator surface →Security & Reliability
Client-owned Azure subscription. Managed Identity. Key Vault. In-region inference and data residency. Aligned with ISO 27001 and ISO 9001. On-premise available.
See the security architecture →Technology
The Machines Like Me App Framework, reference architecture, and the Default vs Private topology options. Under the hood.
See the product under the hood →
See how it works
From one complex process to a running agent.
The path every engagement follows, dramatized in four steps.
Commercial scope. One fixed price per process covers strategy, discovery, build, go-live and a supported hypercare period; then handover to your IT, with an optional support agreement if you prefer not to run it yourselves. If a process proves non-automatable, you receive a voucher for a different one. Payback within a year in the large majority of projects. Full detail in our FAQ →
Describe the problem
“Our order intake takes three people, and every order arrives differently. Can this run on its own?”
Send to Machines Like MeOne process, in your words. That’s the input.
Discovery · process map
- Order captureEvery format, every channelautomatable
- Validation & business rulesMOQs, delivery windows, customer termsdeterministic
- ExceptionsAmbiguity, missing data, edge casesrouted to human
- Integration patternSelected per system during discoverychosen
Discovery-first: the pattern fits the constraint, not the other way around.
Pre-production suite
- Traceable test casesDefined in the specification, reported in the delivery reportpassed
- Scenario testsThe edge cases your team already knowspassed
- Regression testsEvery incident becomes a permanent testpassed
- HITL workflowThresholds and escalation rules declaredconfigured
Nothing ships outside the Governance Envelope.
Order intake agent · run logLive
- 01 · ReceiveNew order received · intake startedqueued
- 02 · ReasonCustomer rules applied · MOQ, delivery windowreasoning
- 03 · CheckConfidence 94% · above thresholdpassed
- 04 · ActOrder posted to ERP · agreed Level 4 step · decision loggeddone
Run by your team. Maintenance available on request.
Get in touch
Talk to us and get started.
Book 30 minutes with our team. Walk us through your process. We'll show you where AI agents can make a difference.