Four stages. Each delivers value on its own. Each makes the next possible.

The first stage is extraction and assembly. biograph works from the design-time artifacts your operation already produces — not from workshops, interviews, or assumptions.

These are not incidental documents. They are the specification of your operating model — created over years, validated through change control, and governing how your operation behaves right now.

For the first time, you can see the operating model as a single entity — rather than as fragments owned by different teams and systems.

The extraction captures what the artifacts actually specify — not what people remember, not what the SOP says if it hasn't been updated, not what the workshop facilitator inferred. The artifacts are the source of truth.

With the specification assembled, biograph reasons over it across systems. This is where the value of the unified model becomes tangible.

Now answerable Why is this MES parameter set to this value? What depends on it in the ERP? Which equipment state transitions constrain the recipe? Where do interface assumptions conflict with how the process actually flows?

These are the questions your team asks in every cross-system meeting — and answers with experience, institutional memory, and educated guesses. No single document has the scope to answer them formally.

The dependencies that span systems are where the real risks live — and where the real optimisation opportunities are.

But they are invisible to any tool that looks at one system at a time. The unified model makes them queryable.

With the specification visible and the dependencies understood, you can address root causes — in the design, not in the execution.

If your operational cycle times are suboptimal, the cause is rarely in how the process is being executed. It is in how the process is designed.

Optimising execution is treating symptoms. Optimising the specification is treating the cause.

Every change is traceable. The reasoning chain — from the process requirement that drives the change, through the dependency analysis that justifies it, to the specific parameter modification — is captured in the model with full provenance. You know what changed, why, what it affects, and what evidence supports the decision.

Working on the specification carries none of the risk of working on the running process. This is not a regulatory event. It is engineering.

In traditional pharma projects, validation is a separate workstream — often consuming as much time and budget as the implementation itself. Requirements traceability is built manually. Test protocols are authored from scratch. Evidence packs are assembled by hand.

biograph inverts this.

The model already knows the specification — what changed, why, the dependencies and constraints. Validation evidence follows directly from the model. Requirements trace to their source. Test cases derive from the specification. Compliance maps to 21 CFR Part 11, EU GMP Annex 11, GAMP 5.

And then The same optimised specification becomes the contextual backbone for what comes next. When you deploy a digital twin, it has a formally specified process to represent. When AI agents move into the execution space, the unified model they need as context already exists.

The data models, integration patterns and system interfaces reflect the actual operating model, with provenance — and they are maintained as the model evolves. Deploying agents without that context is deploying autonomous vehicles onto roads with no maps. biograph builds the map.