How a CoreLine Engagement Starts

Compare workflow, existing-AI and modernisation assessments. See the inputs, decision records, ownership and handover questions to agree before a build.

A conversation, a written scope and a decision

A fit conversation establishes the problem and who owns it. A separately agreed paid assessment can examine the evidence and recommend what to do next. Implementation, a working prototype and ongoing support are included only when they are explicitly part of the written scope.

Before paid work starts, agree the deliverables, required access, fee, schedule, reviewers and acceptance criteria. A discovery report should remain useful if you choose another delivery team. A finding that the prerequisites are missing or that a maintained product fits is a useful result.

Workflow assessment

For an operations or product team with a repeating task and someone accountable for the process.

Bring: a map of the current steps, the systems involved, the exceptions and a few representative cases that may be reviewed. Volumes, handling times and operating costs begin as your assumptions until they can be checked. Agree a safe way to share documents after the initial conversation.

Receive: a process boundary, data and integration assumptions, build/buy/manual alternatives, quality and review criteria, an exception path and a recommended next decision. A first-release outline separates essential work from later scope.

Decide: configure a suitable tool, scope a custom build, improve the process first or stop. A prototype is a separate deliverable if required.

See the illustrative workflow decision record, build-or-buy guide and AI process automation service.

Existing AI assessment

For a team that already has an AI feature or prototype and needs to understand its quality, review effort, latency or running cost.

Bring: the task definition, examples of expected and unacceptable output, the permitted data boundary, current architecture and relevant redacted traces. Include retries, parsing, retrieval and time spent correcting the result. The first public message does not need confidential logs or credentials.

Receive: an evaluation definition and baseline, named failure cases, a comparison of bounded changes and a recommendation. Where data permits, distinguish cost per completed task, model usage, latency and human review. Keep remaining failures visible.

Decide: fix an upstream data problem, change retrieval or validation, evaluate a different model, improve the review path or leave the current approach in place. Release and recovery criteria are agreed before a production change.

Inspect the measured synthetic extraction exercise and AI implementation service. The exercise demonstrates evaluation, not a result on your data.

Modernisation assessment

For a live business application with valuable workflows and parts that have become difficult or risky to change.

Bring: the critical journey, relevant architecture and dependencies, known problems, release process and continuity requirements. Agree read-only access and a safe review boundary where appropriate.

Receive: a keep/repair/replace decision by component, a first milestone, integration and data dependencies, validation criteria and a recovery plan. Any small fixes included during assessment are stated separately in the scope.

Decide: improve an existing component, transition a bounded workflow or investigate a prerequisite before replacing it. Reverting application traffic does not automatically reverse a database migration.

Read the filled architecture decision example and legacy modernisation service.

Ownership, access and delivery decisions

DecisionWhat to agree in the engagement
Source code and IPOwnership, repository access and any third-party licence obligations.
InfrastructureWho owns the accounts, approves access and pays operating charges.
DataWhat may be processed, where, by which service and under whose approval.
AcceptanceRepresentative cases, responsible reviewers and the evidence needed to release.
Production changeWho approves release, monitors the result and can invoke recovery.
HandoverDocumentation, runbooks, access transfer and any training included.
SupportCoverage, response expectations, exclusions and responsibility after launch.
Commercial scopeFee, schedule, assumptions, changes and the point at which a new commitment is needed.

These are decisions to document, not a substitute for the signed agreement or a promise of a particular support level. We confirm project-specific terms before committing to them.

Prepare for the first conversation

Describe one problem and what a useful result would enable. Mention the systems, the person responsible and any real deadline. If you already have a proposal, evaluation or architecture sketch, say that it exists. We can agree what to review and a safe channel for sharing it.

You do not need to choose a model, write a full RFP or upload production records before we establish the fit.