All demos

AI Powered Tools · Meridian Office Group

AI Powered Tools demo - Meridian Office Group

Hours lost to reading, sorting, summarising and answering the same things by hand.

Meridian Office Group - Document Review Assistant

Honesty note. This is a static demonstration using mock AI outputs. Nothing here calls a live model - every summary, field and draft is pre-written. In a real build, outputs would be connected to an approved AI model or workflow, with a person reviewing what matters.

Explore freely: switch documents, generate summaries, extract fields, show the source, edit the draft, tick the review checklist, or clear the output and start again.

Incoming documents

Triaged by review status

Supplier Agreement - Northstar Components

Meridian Office Group & Northstar Components Ltd

Medium risk
Type
Contract
Pages
11
Received
18 Jun 2026
Owner
Procurement

Source text (mock)

  • p.28.1 This Agreement shall run for an initial term of twelve (12) months and shall renew automatically for further twelve (12) month terms unless either party gives written notice at least sixty (60) days before the renewal date of 31 October 2026.
  • p.612.4 The parties shall comply with applicable data protection law. [No data-processing appendix is attached to this Agreement.]
  • p.33.2 The total annual charge payable by Meridian Office Group shall be £42,000 (forty-two thousand pounds), invoiced quarterly.
  • p.714.2 The total aggregate liability of the Supplier under this Agreement shall not exceed the total charges paid in the preceding twelve (12) months.
  • p.1Account owner (Meridian): Procurement team - supplier relationship managed by the Procurement function.

AI summary

Turn this 11-page contract into a short, checkable summary.

Extracted fields

Pull the key values out of the document into a list you can check.

Support assistant

Pick a question. Answers come from approved information, with the source shown.

Mock answers from a fixed set - a real assistant would search your approved sources.

Draft reply

Suggested next step: Ask the supplier to confirm whether personal data is processed under this contract, and request a data-processing appendix if it is.

Human review checklist

The Approve button stays locked until a person has checked everything. The tool never approves itself.

Review checks before approval

4 checks left before you can approve.

The outcome. AI speeds up the repetitive parts - summarising, extracting and drafting - so a person still reviews and approves every document before anything is acted on.

Read this demo as plain text

The Document Review Assistant lists incoming documents (contracts, invoices, policies and enquiries) for Meridian Office Group. Selecting one - for example the Supplier Agreement with Northstar Components - loads its metadata and source text into a viewer.

A person can generate a pre-written summary, extract key fields (value £42,000, renewal date 31 October 2026, notice deadline 1 September 2026, owner Procurement, overall risk medium), and trace each value back to the exact source line. Risk flags include an auto-renewal clause, a liability cap at the contract value, and a likely missing data-processing appendix.

A draft supplier question can be generated and edited. Before anything is approved, a human review checklist must be completed: read the summary, confirm the fields, review the risk flags, and assign an owner. Only then does the Approve button unlock. Every AI output is mock data; the tool summarises, extracts and drafts but never makes the decision.

What this demonstrates

  • AI turns a long document into a short summary and structured fields a person can check.
  • Every extracted value links back to a source line, so claims are easy to verify.
  • Drafts are generated as a starting point and stay fully editable before anything is sent.
  • An approval is locked until a human review checklist is complete - the tool never decides on its own.

Popular AI tools I can build

Pick one to see who it is for, what it adds and what it could include.

Contract summary assistant

Medium to high

Summarise long agreements and surface key dates, values and risk points for a person to review and confirm.

Typical users
Procurement, operations and legal-adjacent teams
Value added
Faster first-pass review, with a human confirming every figure.

Could include

  • Document upload
  • Summary
  • Field extraction
  • Source linking
  • Review step

Want to use AI on a real task safely?

You do not need a perfect brief. A messy process is enough to start.

Discuss a projectTalk through an idea

Next demo

Compliance - Records that are easy to prove.

Open compliance demo
What a real build could include · read as plain text

What a real build could include

The pieces a production version would be assembled from, in plain terms.

  • Document handling

    Upload, list and view the source material in one place.

  • Summarisation

    Condense long text into a short, checkable summary.

  • Field extraction

    Pull structured values like dates and amounts, each with a confidence note.

  • Source linking

    Show the exact line in the document each output came from.

  • Assistant Q&A

    Answer common questions from approved information, with sources.

  • Drafting

    Generate editable replies and content as a starting point.

  • Human review

    A checklist and approval step gate anything before it is used.

Read plain explanation

The situation

Meridian Office Group (Procurement, finance and office services handling contracts, invoices and enquiries). Hours are lost reading, sorting, summarising and answering the same documents by hand. Long agreements are read line by line, the same questions get re-answered, and key dates are easy to miss in a busy inbox. This demo shows how ai can summarise, extract and draft to speed up the repetitive parts, while a person reviews and stays responsible for every decision.

How it would be built

  1. Target. Pick a specific, repetitive task worth automating - here, triaging incoming documents.
  2. Apply. AI summarises, extracts and drafts only where it genuinely saves time.
  3. Control. A person reviews the output, checks the source, and approves. The decision stays human.

What the demo points out

  • 01. Document helper. Key fields pulled out of a long supplier agreement into a checkable list. AI is useful where it turns long documents into structured fields that people can check. Value: Faster review.
  • 02. Support assistant. An assistant answering common questions from approved information, with sources. Common questions can be answered from approved information instead of repeated manually. Value: Consistent answers.
  • 03. Content system. An editable draft reply generated from the document and its flags. Drafts can be generated quickly, but remain editable and reviewable. Value: Faster writing.
  • 04. Human control. A review checklist that must be completed before the Approve button unlocks. The tool does not replace judgement. It speeds up the repetitive parts and keeps the person responsible. Value: Safer AI use.
  • 05. Source linking. Each extracted value can highlight the exact mocked line it came from. AI output is only trustworthy when you can trace it back to the original text. Value: Checkable claims.
  • 06. Inbox triage. Incoming documents sorted by type, owner and review status. Sorting and routing the inbox is repetitive work AI can handle before a person looks. Value: Nothing slips.

What this demonstrates

  • AI turns a long document into a short summary and structured fields a person can check.
  • Every extracted value links back to a source line, so claims are easy to verify.
  • Drafts are generated as a starting point and stay fully editable before anything is sent.
  • An approval is locked until a human review checklist is complete - the tool never decides on its own.

Want to use AI on a real task safely?