Build Your
AI Operating System
Building a shared AI Platform
Control access
Connect approved tools
Track cost and quality
Teams share one controlled way to use AI.
Where the AI in your business actually lives
Have your sales, marketing, ops, and finance teams started using AI? Chances are each is using different copilots and automation tools connected to AI models, some of which may be through accounts you’ve not vetted. This can be a challenge to manage from an IT perspective, as it’s unclear what data employees are sharing, what happens to it, what models are being used, and the costs are hard to oversee. We build a platform that sits beneath the tools they’re using, making sure only approved models are being used, giving insight into what’s happening, and controlling who has access to what, all without disrupting their workflows.
What does it entail?
Inventory: we’ll review what’s being used across the business and put together a report on tools in use, a rough estimate of who’s using them, what data they have access to, and associated risks.
Platform: the platform itself has several layers:
Governance over who has access to the various elements
A common memory of transcripts, documents, data, etc., either already available in the organisation or connected, that employees have authorised AI agents to use
A router for accessing various LLMs that optimises for cost/performance based on the task at hand
Various tools that AI agents can be given access to in order to carry out work, such as read/write to certain systems, access certain information, create presentations
Guardrails to make sure they’re not saying or doing anything you don’t want them to.
All activity will be logged, including what was requested, who requested it, which model it went to, and what the response was. Costs will be logged by department to enable oversight of spend.
AI use policy: this can be written/updated as part of this exercise, ensuring it passes muster with your board.
How does this help?
For your engineering/IT teams it offers control over who has access to what models/data/tools, insight into what’s going on, and oversight of costs.
For employees, it allows them to continue using tools they’ve already adopted, possibly with better models/tools/memory to improve performance.
For the organisation, it helps prepare for AI use becoming more regulated: the EU AI Act already has some obligations and more are coming for high-risk uses. It would be preferable to be ahead of this, rather than rushed into doing so when a client/insurer/etc. asks.
How do you carry this out?
We’d start with a call with relevant technical decision makers to understand what’s been implemented, any particular requirements, etc. Then, if desired, carry out the above exercise:
- Agree scope of audit, agree deliverable format, carry out audit, deliver results
- Agree scope of AI policy update, agree deliverable format, write/update policy, deliver results. This will ensure it covers any obligations from ISO/IEC 42001, which is useful even though you won’t be certified as this requires an external review, AI use cases within High-risk category of the EU AI Act and others if desired.
- Agree desired functionality, agree deliverable format, build out solution in phases, test, deploy, deliver. This will be built in your cloud account, repos, following your architecture. We can help with this if you don’t have these setup. We’d recommend using various open-source tools and configuring them rather than building from scratch. We’ll hand over all code, but can help with changes if desired.
- Agree format of employee training, carry out training. We can help with employee communications around this change if desired.
- Agree scope of AI use documentation which may be included as part of policy, agree deliverable format, write documentation, deliver results.
- Agree format of tech handover documentation, agree on-call rota if applicable, agree process for future changes and eval suite for releases, document, and deliver. We can continue to maintain the tooling on a retainer basis if desired.
Book a call
Pricing will be fixed rate for the audit and AI policy update, quoted price for each phase of the shared AI Platform. Savings come from removing departmental AI subscriptions, optimising LLM spend, and saving time on compliance.
Book a call here.