RSIAI Get in touch

Development / Public roadmap

Progress that follows
the evidence.

A proposed twelve-week path from a concrete problem to an informed pilot decision. Each stage has an acceptance gate, not an automatic promise of a launch.

V1.1 / 7 OCT 2026200 PLANNED HOURSSEQUENCE SUBJECT TO CUSTOMER ACCESS
Available now: browser-based interactive prototypes for enquiry-to-next-action and token optimisation, a secondary synthetic evidence workflow, the public whitepaper and a local pilot-brief builder. The twelve-week plan would validate one customer workflow. Live model processing, external scheduling or quotes, customer integrations and measured outcomes are future work.
Weeks 1–2Planned · 32 hours

Find the repeated problem.

Interview operators, map one recurring task and test the synthetic prototypes.

Evidence to proceed: A process owner, a recurring problem and a clear gap in the current workflow.
Implementation milestones
  • Observe a recent enquiry-to-next-action or evidence-handover case.
  • Record handling time and existing software capability.
  • Agree what a useful result would look like.
Weeks 3–4Planned · 35 hours

Define the smallest useful pilot.

Agree authorised samples, acceptance measures and a commercial decision point. Establish the private application foundations.

Evidence to proceed: Sample permission, an agreed scope and credible willingness to pay.
Implementation milestones
  • Specify allowed documents and retention.
  • Establish tenant and storage boundaries.
  • Prepare one scoped statement of work.
Weeks 5–6Planned · 35 hours

Make the checks dependable.

Implement the chosen workflow, approved rules, relevant source references and spending limits.

Evidence to proceed: Missing inputs and conflicting rules stay visible; attempts and spending remain bounded.
Implementation milestones
  • Version rules, documents and model configuration.
  • Measure quality on an initial fixed set.
  • Reserve budget before each model attempt.
Weeks 7–8Planned · 35 hours

Keep review and action connected.

Build correction, source-bound approval and export. Add one connector only where the customer needs it.

Evidence to proceed: Changes invalidate approval; duplicate and uncertain external actions are recoverable.
Implementation milestones
  • Preserve a reviewer’s edits.
  • Keep the workflow decision separate from delivery status.
  • Test interruption, replay and approval changes.
Weeks 9–10Planned · 33 hours

Test it against the baseline.

Evaluate quality, access isolation and recovery. Run a controlled customer pilot where agreed.

Evidence to proceed: Comparable results for quality, staff effort and cost per accepted workflow outcome.
Implementation milestones
  • Use a separate holdout case set.
  • Measure corrections and exceptions as part of handling time.
  • Verify restore and deletion procedures.
Weeks 11–12Planned · 30 hours

Make an evidence-based decision.

Compare outcomes, fix the remaining material issues and decide to continue, revise or stop.

Evidence to proceed: A useful result, viable total cost and a real customer decision.
Implementation milestones
  • Record limitations alongside results.
  • Agree ongoing service only when value supports it.
  • Reuse the validated components in the next implementation.

The week-four decision

Demand is a requirement.

If there is no specific process, authorised sample access or credible willingness to pay, the next step is to revise one adjacent workflow and repeat discovery.

Beyond the first pilot

Repeatability comes next.

Later work may include configurable industry templates, additional integrations and more advanced usage optimisation. Expansion depends on delivery evidence and support capacity.

A public-sector path, when ready.

Future service-delivery work would require a suitable problem, agency-specific procurement and data handling, delivery evidence and an accountable operating model.

Research & funding pathways