Free shadow-AI scan
Upload your supplier spend.
See your real AI bill.
Export supplier payments from Xero, QuickBooks or NetSuite, upload the file, and see every AI vendor in your ledger - including the ones nobody told IT about. No integration, nothing installed, nothing stored. Results in minutes.
2 unknown to IT
£2,370/month of AI spend with no owner, no measurement, and a renewal nobody is watching.
How AIchemi works
From free scan to accountable AI programme.
Free Scan
Upload a supplier-spend export. See every AI vendor in your ledger, including shadow spend. Headline free; a free account unlocks the full report.
FreeSelf-Service Assessment
Alfred runs a two-week AI-ROI assessment - interviews, measurement, and a board-ready pack. Under two hours of your time.
One-offOn-Going Self-Service
The meter stays on: verdicts refresh monthly, a weekly digest, market alerts. Assessment fee credited.
RecurringCustom (Human Included)
Optional human help - training your internal AI team, leadership workshops. Priced, scheduled, capped.
CustomThe product - the Audit
A consultant-grade AI audit, run by AI.
Alfred - our AI advisor, built on a working consulting methodology - conducts the whole engagement: it interviews your department heads, reads your ledger, measures what’s really happening, and writes the board pack. Humans are optional, and billable. Only 22% of finance leaders can tie AI spend to business outcomes, and 66% of boards now condition further AI funding on proof of return (“Finding the ROI of AI,” CloudZero, June 2026) - the audit is how you join the 22%.
Runs the engagement
Intake interview, data checklist, department-head interviews via simple links (no accounts needed), analysis, report. You’re never asked to do more than one thing at a time.
The board pack
Where your AI money actually goes (shadow spend included), a verdict per department with the evidence attached, an hours-saved value model, and a 30-60-90 plan.
≤ 2 hours of your time
Plus 15 minutes per department head. Two weeks elapsed. Fixed price, paid by card - no procurement expedition.
The data-gaps appendix. Every pack ends with what the audit couldn’t know and exactly what would close each gap. When the data can’t support a verdict, the pack says “unknown” - it never fabricates a number to look finished. That’s why your board can trust the pages that do have numbers.
How it works - a real walkthrough
Two weeks, start to boardroom.
Maya is ops director at a 140-person firm. Her CFO has asked “what are we getting from all this AI?” two quarters running.
The free scan
Maya exports last quarter’s AP lines from Xero and uploads them. No signup for the headline: 7 AI vendors, £5.2k/month - two of them unknown to IT. She forwards the report to the CFO.
“OK - what’s the full picture?” - the CFO, four minutes later
She buys the audit
Fixed price, by card, under her approval limit. Alfred’s intake interview takes 20 minutes: departments, the tools she knows about, what the board actually asked.
Data, one item at a time
Alfred’s checklist: org chart CSV from HR, FTE cost bands, the OpenAI account IT will actually sign off. She skips the rest - Alfred says plainly what skipping costs, and files it for the data-gaps appendix.
Alfred interviews her department heads
Five heads, one link each, 15 minutes per person - what they use AI for, where the time really goes, what’s friction. Two ignore the first email; Alfred nudges once and tells Maya about the third. Everything captured queues for her confirmation - nothing enters the record without her sign-off.
The board pack lands
Spend map with the two shadow vendors flagged. Marketing: working, with evidence. Engineering: at risk - spend up, mapped value flat. Finance: honestly unknown, with exactly what would make it knowable. A 30-60-90 plan. Maya presents it at the next board meeting mostly as-is.
The meter stays on
The pack’s closing page: the measurement system is already installed - a subscription keeps verdicts fresh monthly, audit fee credited. The CFO, who has now seen evidence instead of vibes, approves. Three months later Maya books a white-glove enablement day for the two teams that want to build their own use cases.
The design rule behind this flow: the customer never does more than one thing at a time, never waits without knowing why, and never hears “unknown” without hearing what would fix it.
The platform underneath
Three data streams. One honest picture.
ROI has three ingredients - who is using AI, what they’re doing with it, and what it costs. Most tools see one. AIchemi connects all three, and Alfred reads them together.
Your people & departments
Live from your identity or HR system - or a simple CSV. A human review queue guards every change.
- SCIM 2.0 provisioning
- HR systems via one unified integration
- Joiners, movers and leavers tracked
Your AI platforms
Usage and spend synced from the tools your teams actually use, per department.
- Connectors for five major AI platforms
- Scheduled daily sync
- Department-level heatmaps
Your finance system
Line-level AP data from NetSuite, QuickBooks or Xero - or the CSV you already exported for the scan.
- Shadow-AI detection from invoices
- Department coding preserved
- Never mixes estimates with actuals
Meet Alfred.
Your AI advisor reads all three streams and delivers judgments, not dashboards: department verdicts each month, goal plans with concrete steps, a weekly digest, and a conversation you can push back on. Alfred remembers what worked at your organisation and gets sharper with every cycle.
Security & Data
Choose where your data lives - and who your AI runs on.
We publish exactly what we do today and what’s on the roadmap. No badge we can’t defend in a security review.
Standard
Managed cloud, UK-resident. Your database lives in London (AWS eu-west-2), encrypted at rest (AES-256) and in transit (TLS 1.2+), with per-organisation row-level isolation enforced in the database itself. Your data is downloadable in one click, subscribed or not.
All plans - available nowBring your own AI
Alfred’s analysis runs on your AI provider account - Anthropic, OpenAI or Google - under your key, your data-processing agreement, your retention terms. If your key is ever unreadable, the system fails closed: it stops rather than silently rerouting your data to ours.
Available nowYour cloud
Deploy AIchemi’s data plane in your own AWS, Azure or Google Cloud tenancy - your infrastructure, your region, your compliance regime. Combined with bring-your-own-AI, nothing of ours touches your data. Includes EU-sovereign configurations. We’re building this with our first enterprise partners.
Enterprise roadmap - talk to usWhat we do today
- Row-level security scoping every table to your organisation
- All credentials encrypted at rest; inbound tokens stored as hashes
- Individual-level usage masked by default - admins see teams, not people
- Per-organisation AI budget caps and full usage metering
- One-click data takeout, subscribed or not
- Adversarial security review as part of every release
Subprocessors
- Supabase (database - AWS London)
- Vercel (application hosting)
- Anthropic (managed AI analysis - unless you bring your own)
- Stripe (payments)
- Merge (HR & accounting sync)
- Sentry (error monitoring)
- Tavily & Google (market-alert scanning - no customer data)
- Resend (email - coming soon)
Pricing
Start free. Pay for the audit. Keep the meter running.
Priced per organisation, not per seat - seat pricing punishes exactly the adoption we’re trying to grow.
Free Scan
£0Upload a supplier-spend export, see your real AI spend in minutes. Headline free; a free account unlocks the full report.
Scan nowSelf-Service Assessment
Founding pricing one-offThe full Alfred-led engagement: interviews, measurement, board pack. Two weeks, ≤2 hours of your time.
Start an assessmentOn-Going Self-Service
Founding pricing /monthVerdicts refresh monthly, weekly digest, market alerts, goals. Assessment fee credited to year one.
Keep the meter onCustom (Human Included)
Per engagementHuman help: train your internal AI team, leadership workshops. Delivered by the founders on a capped calendar.
See servicesAbout
Consultants who got tired of un-measurable advice.
AIchemi encodes a working AI-adoption methodology into software - so the advice arrives with the evidence attached, and every engagement makes the methodology sharper.