The AI Touchstone suite
Enterprise AI spend lacks the certainty of success
Successfully moving to an AI enabled business model relies on several key stages. Yet, with all the focus on the models, there are few helpful tools and methods to support success.
Without the ability to build, register and reuse applications, calculate ROI, and monitor and control costs, there is no guarantee of success or demonstrable business value.
The AI Touchstone suite of products will help you successfully use generative AI and agents for operational efficiencies and customer solutions.
See below for more details on the individual modules.
The platform
One accountability stack, from first experiment to post-production governance
Each module covers one stage of an AI application’s life. Together they keep cost, ownership and value connected to the same record, from the first prototype to the day an agent is retired.
- Stage 1 · Experiment
Foundry
Business subject matter experts build agents in a secure, multi-model workspace.
- Stage 2 · Productionise
App Registry
Proven applications are registered, owned and reused.
- Stage 3 · Prove value
Attribution Engine
Spend is tied to validated business impact.
- Stage 4 · Operate
Operations Dashboard
Every model call is metered as a financial unit.
- Across every stage · Govern
Assurance layer
Honest confidence, audit trail and security controls.
1Foundry
Create agents and applications without waiting.
Enable subject matter experts to create agents and applications themselves, without waiting for an overstretched technology team, delivering quicker and more effective results.
Foundry ensures promising ideas start delivering measurable enterprise value quickly.
Secure environment
Experiments run in an isolated workspace, inside your own controls.
Multiple model support
Choose the right model for each task to ensure effective usage.
Instant review and publication
Agents and applications pass review and publish in one step.
Foundry addresses the gap between an idea and a governed working solution.
- Step 1Discover
- Step 2Provision
- Step 3Pilot
- Step 4Scale
Illustrative product preview · sample data
2App Registry
Take proven applications from prototype to production.
Without the ability to move a project from prototype to production, the work risks becoming a science project.
Enabling others within the organisation to reuse and potentially adapt proven applications brings economies of scale.
Listing
Every agent registered with owner, version and workflow.
Policy and access
Scoped permissions, revoked the moment an agent goes idle.
Regularising and decommissioning
Dormant agents retired on schedule.
One registry to manage, deploy and govern AI agents throughout their lifecycle.
| Agent | Owner | Access | Status | Spend |
|---|---|---|---|---|
| invoice-triage-agentv3.2 · finance-ops | Priya K. | Scoped | Active | $41,208 |
| support-copilotv5.0 · customer-care | Marcus D. | Scoped | Active | $28,940 |
| legacy-summarizerv1.4 · mcp | No owner | Revoked | Orphaned | $6,104 |
| q3-forecast-botv0.9 · planning | Aisha R. | Revoked | Deprecated | $912 |
Illustrative product preview · sample data
3Attribution Engine
Demonstrate the business impact of AI.
Demonstrating relevant business impact is critical to maintaining funding, planning budgets and prioritising where the greatest efficiency gains can be made.
Multiple frameworks
Templates that express impact in terms each function recognises.
Credited by contributor weight
Value is divided between contributors, with no double counting.
Traceable totals
Every total relates back to a source record that can be viewed.
A framework driven model associates application expenditure directly with validated business impact.
- support-copilot: 62%
- KB search MCP: 23%
- Routing agent: 15%
SR-2291Ticketing system export, 14,208 rowsIllustrative product preview · sample data
4Operations Dashboard
Gain complete visibility into AI execution expenses.
Detailed telemetry records model requests as individual financial units.
Track actual or forecast charges by provider, specific model and route, providing unified financial clarity across all active tools, agents and protocol integrations.
Every call is a span
Each agent call is logged on its own, never blended silently.
Real-time roll-ups
Spans roll up by agent, tool and MCP as they arrive.
Attributable to source
Full visibility of your spend, down to the request that caused it.
Granular financial tracking records model interactions, service calls and protocol tasks precisely.
Fed by
↓ Span ledger · one record per request
Illustrative product preview · sample data
◆Assurance layer, post-production governance
Never one green number.
The cost is exact. The value carries its confidence. That honesty is the product.
Illustrative product controls and audit entries.
The honesty primitive
Facts and estimates are labelled differently everywhere they appear.
A fact, read straight from your system of record.
A modelled estimate, always shown with its confidence.
Hash-chained audit trail
Every change to an agent, a policy or a figure is appended, never overwritten.
prev 9f3c…a71e → 2b04…d913prev 2b04…d913 → c7e8…0f42prev c7e8…0f42 → 51aa…6be0Built security-first
Controls that let security and procurement teams sign off.
- SSO & SCIM ready On
- Tenant-isolated data On
- No training on your data On
- Hash-chained audit trail On
The ones who survive 2027 will be the ones who could prove it.
Early design partners help shape the instrument, before it meets a roadmap.