Where does our data go?

MLX connects read-only to the systems you approve and keeps each target scoped to your organisation. Those source systems remain authoritative. When a workflow uses an external model provider, the context needed for that request follows the provider path approved for your deployment.

Do we need to migrate our legacy systems first?

No wholesale or rip-and-replace migration is required. A focused activation connects the priority systems you approve, starting read-only, and brings the relevant evidence into the Twin while those systems remain authoritative. Mercury Labs engineers then work with your experts to turn that evidence into reviewed business context.

Can we choose models and control their cost?

Yes. The approved business context stays separate from the model provider. You can use stronger models for demanding work, lower-cost models for routine tasks or evaluate another configured provider without rebuilding that context. Material changes are tested before rollout.

MLX is a scoped engagement. Mercury Labs engineers support the activation and operate the configured target, while model usage is recorded where provider evidence is available. Talk to us for a proposal based on the systems, workflows and operating requirements in scope.

Why not just use ChatGPT or Copilot?

Use them where they fit. The Twin gives approved AI tools the reviewed business context they normally lack, including agreed definitions, source freshness and evidence about what each answer used. It complements the AI surface rather than replacing it.

We're regulated. Can we actually use this?

MLX starts scoped to your organisation and read-only, with human review before business context is published. Normal governed queries do not include browser, terminal or write access. A regulated deployment still requires approval of the exact data and model paths, evidence retention and operating controls.

Which systems can you connect?

Current connectors: Xero, SunSystems, Stripe, GoCardless, Microsoft 365 and SharePoint, Google Drive, Docs and Sheets, Gmail, Google Calendar, Outlook, Slack, Microsoft Teams, Granola, Notion, Linear and GitHub. Legacy ledgers such as Infor SunSystems stay queryable without migration; exact scope is confirmed before activation.

What happens when the Twin cannot support a question?

It says so. A governed answer identifies the approved product, version and freshness it used. If that product does not support the requested question or evidence, the Twin returns 'not supported' instead of producing a plausible guess.

Drawing 00 — The AI Twin

Your business, twinned for AI.

Generic AI doesn't know your business. MLX constructs a governed Twin from the systems you already run—reviewed by your team and usable by the AI you approve.

Discuss your first activation See how it works

Front elevation — The AI Twin

Twin process: approved sources connect, relevant evidence gathers, business context is proposed, and your team reviews and approves it.

Connected evidence sourcesYour source systems remain authoritative

Accounting · payments

Xero

SunSystems

Stripe

GoCardless

Documents · files

Microsoft 365

SharePoint

Google Drive

Docs · knowledge

Google Docs

Google Sheets

Notion

Email · calendar

Gmail

Calendar

Outlook

Chat · meetings

Slack

Teams

Granola

Engineering · projects

GitHub

Linear

Further sources scoped per activation

MLX: One trusted view of your business
Built from: Your systems, terms, rules and ways of working
Evidence: Approved sources stay authoritative
Approved by: Your team, before AI relies on it
Used by: The AI tools and workflows you approve
Model choice: Use the best configured model for each task · switch without rebuilding the context
Operated: Managed and governed in MLX

01 — The missing context

AI has intelligence. It does not know your business.

Generic AI knows the world. It does not know what your accounts mean, how work moves, or which rules your team trusts. Every AI pilot rebuilds that context from scratch, and answers arrive unverifiable.

An AI Twin closes that gap. It turns the evidence in your systems into reviewed, versioned business context the AI tools you approve can query.

Why now

The model supplies intelligence. The Twin supplies your approved business context.

Modern and legacy systems

Make legacy data usable by AI—without replacing the systems that hold it.

MLX connects the priority sources you approve and turns their records into reviewed business context AI can use. Your existing systems remain authoritative, so you can start with legacy applications, specialist databases, spreadsheets and documents without a wholesale migration.

02 — How the Twin forms

One business. Three states you can inspect.

You can see what the Twin read, what it derived and what your team approved. Source evidence, transformation and the governed result remain distinct and inspectable.

01
Connect source evidence
Scoped, read-only connectors bring relevant evidence from approved source systems into the Twin while those systems remain authoritative.
02
Prepare business context
MLX proposes definitions and calculations, links them to their sources and records each approved version.
03
Review, approve and publish
Your people review the proposals. What they approve becomes durable business context that approved AI can use.

Exhibit — the governed result, as an agent receives itIllustrative specimen

Question

“What was revenue last quarter?”

A generic copilot

“Revenue was around £1.2m, based on the documents I could reach.”

SourceWhatever it could reach

VersionNone

FreshnessUnknown

UnsupportedAnswers anyway

An approved Twin answer

£1,243,908 · Apr–Jun 2026

Approved definitionMonthly revenue · v2

Data current to30 Jun 2026

Evidence usedPublished business view

If unsupportedSays so · never guesses

Prepared business logic

Fast answers from approved calculations.

MLX prepares approved business calculations before AI needs them. That produces significantly faster answers, consistent numbers and lower model costs.

When the approved calculation and source data have not changed, the result does not change either—and its evidence stays attached.

03 — Why a Twin

One trusted business context. Many approved AI tools.

A governed Twin gives the AI you approve one shared, explainable view of how your company works. Your systems remain authoritative, and your people decide what becomes trusted context.

01
One trusted business context
Approved AI tools work from the same versioned understanding of your business instead of rebuilding it in every chat and workflow.
02
Use the data you already have
Make modern and legacy data usable by AI without replacing the systems that hold it. Those systems remain authoritative.
03
Answers show what they used
Each governed answer identifies the approved definition and version it used, with governed evidence references. Configured products can also expose source freshness.
04
Choose models and control cost
Use stronger models for demanding work, lower-cost models for routine tasks or change providers without rebuilding the Twin's business context.

Exhibit — one Twin, approved AIKeep the map · choose the intelligence

Approved context · available through configured AI tools

MLX
ChatGPT
Claude
Gemini

One governed Twin · approved AI

One Twin. The right AI for each job.

Use the same approved business context through MLX and the AI tools configured for your deployment. Each one works from the same reviewed understanding of how your business operates.

04 — Proof from real systems

From existing systems to trusted results.

MLX connects to the systems you already run, reconstructs the definitions and calculations behind your numbers, and gives your team results they can inspect, correct and approve. These examples come from real modern and legacy finance environments.

01

Accurate and explainable

Results your team can verify.

The Twin produced a P&L from reviewed business definitions and calculations. When finance checked it line by line against its approved ground truth, all nine lines matched, with £0.00 variance. The definitions, calculation version and supporting evidence remain traceable.

DrawingMLX-PROOF-001

MethodChecked line by line

StatusFinance verified

To the
penny

Revenue£2,480,000.00£2,480,000.00£0.00

Cost of sales£(1,041,600.00)£(1,041,600.00)£0.00

Gross profit£1,438,400.00£1,438,400.00£0.00

Admin expenses£(612,300.00)£(612,300.00)£0.00

Payroll£(548,900.00)£(548,900.00)£0.00

Depreciation£(72,400.00)£(72,400.00)£0.00

Operating profit£204,800.00£204,800.00£0.00

Taxation£(51,200.00)£(51,200.00)£0.00

Net profit£153,600.00£153,600.00£0.00

Checked line by line · 9 of 9 lines matchedLargest variance £0.00

Representative figures protect confidentiality · measured result: 9 of 9 lines matched

02

Modern and legacy systems

Works with the systems you already run.

MLX is already working with data from Xero and a legacy SunSystems ledger. Both systems remain authoritative while the Twin turns their data into reviewed metrics and processes—without a wholesale migration.

Modern cloud accountingConnected · Xero

Read-only connection

Business logic prepared

Gross margin 58.0%, calculated from Revenue ▸ Cost of sales

Source remains authoritativeTeam reviewed

Legacy accounting systemConnected · SunSystems

Read-only connection

Business logic prepared

Aged creditors £412,806.00, calculated from Ledger ▸ Purchases

Source remains authoritativeTeam reviewed

Real operating environments · read-only connections · reviewed by the team

03

A focused first result

Start with something useful—not a replacement programme.

A focused two-week activation starts with one priority workflow and the systems behind it. Your team reviews the first business context before deciding where the Twin should expand next.

05 — Activation

Start with one useful part of your business.

Connect priority systems, review what MLX constructs, then put the approved context to work in a focused two-week activation.

01
Connect priority systems
Connect the systems that matter first, starting read-only. MLX identifies the evidence they can support.
Days 1–5
02
Review what MLX proposes
Your experts inspect the proposed definitions, relationships and measures, resolve exceptions and approve what is reliable.
Days 6–10
03
Put approved context to work
Publish the approved business context and use it through the AI tools configured for your deployment.
End of week 2

06 — Buyer questions

What you control, and how it works.

Clear answers about data, model choice, deployment, cost and governance—grounded in the product available today.

MLX connects read-only to the systems you approve and keeps each target scoped to your organisation. Those source systems remain authoritative. When a workflow uses an external model provider, the context needed for that request follows the provider path approved for your deployment.

07 — Your first activation

Give your AI the business context it lacks.

Choose a priority workflow, connect the systems behind it and review the first business context before approved AI puts it to work.

Discuss your first activation →Mercury Labs

Product

The AI Twin

Deploy

Managed targets

Built by

Mercury Labs · London

Your business, twinned for AI.

AI has intelligence. It does not know your business.

Make legacy data usable by AI—without replacing the systems that hold it.

One business. Three states you can inspect.

Connect source evidence

Prepare business context

Review, approve and publish

Fast answers from approved calculations.

One trusted business context. Many approved AI tools.

One trusted business context

Use the data you already have

Answers show what they used

Choose models and control cost