Andrew AI

We turned a 30-year-old paper-based accountancy firm into an AI-native business where AI agents handle operations, allowing accountants to serve more clients at a fraction of the cost.

Comment

Inspiration

Andrew AI started with a real business, not a hypothetical problem.

For over 30 years, Andrew & Co has provided accountancy services to small businesses in London, today serving around 100 clients.

Yet much of the firm's work still relied on emails, spreadsheets, paper documents and repetitive manual administration. Accountants spent valuable time collecting documents, identifying transactions, updating working papers, chasing clients and checking whether information was missing.

We asked a simple question:

What would happen if we rebuilt a traditional accountancy practice around Gemini?

The goal was not to replace the accountant. It was to give the accountant an AI-powered operating system that handles repetitive operational work while keeping humans responsible for review, approval and professional judgement.

That became Andrew AI.

What it does

Andrew AI transforms the day-to-day operation of an accountancy practice into an intelligent, automated workflow.

Instead of accountants manually moving information between emails, documents and spreadsheets, Andrew AI helps manage the process from document intake to accountant review.

Andrew AI can:

  • Automatically ingest client documents and attachments
  • Use Google Gemini to understand and classify financial documents
  • Extract structured financial information from unstructured documents
  • Organize information by client and financial year
  • Detect missing client documentation
  • Generate intelligent document requests and reminders
  • Maintain structured digital working papers
  • Give accountants a central dashboard showing the status of every client
  • Track processing failures and items requiring human review
  • Maintain an audit trail showing what happened, when, and by whom
  • Support secure team workflows with clear accountability

The accountant remains in control of the process.

Gemini does the repetitive work. Humans provide oversight, judgement and approval.

A real-world workflow

Andrew AI was designed around the workflows of an actual accountancy practice.

A typical workflow looks like this:

1. Documents arrive

A client sends financial documents or bank statements to the accountancy practice.

↓

2. Andrew AI processes them

The system identifies the client and uses AI to understand and classify the documents.

↓

3. Gemini extracts the information

Gemini converts unstructured financial documents into structured information that can be used in the accounting workflow.

↓

4. Working papers are updated

Relevant information is organized into the client's digital working papers.

↓

5. Missing information is detected

Andrew AI identifies gaps in the client's records and can prepare or send appropriate requests.

↓

6. The accountant reviews

Items requiring professional judgement remain visible for human review and approval.

This turns what was previously a fragmented manual process into a continuously operating AI-assisted workflow.

How we built it

Andrew AI combines Google Gemini with Google Cloud and Google Workspace technologies to connect AI reasoning directly to real accounting workflows.

Our stack includes:

  • Google Gemini 2.5
  • Vertex AI
  • Google Cloud
  • Google Cloud Run
  • Google Sheets
  • Gmail integrations
  • Next.js
  • TypeScript
  • PostgreSQL
  • Supabase
  • n8n
  • REST APIs
  • Document processing and OCR

Gemini is not simply being used as a chatbot.

It acts as an intelligence layer inside the accounting workflow, interpreting documents, extracting information, classifying data and helping determine what should happen next.

Google Sheets provides familiar structured working papers that accountants can continue to inspect and work with, while Andrew AI orchestrates the surrounding workflow.

This was important to us because successful AI adoption should not require professionals to abandon every tool and process they already understand.

Human-in-the-loop AI

Accounting involves sensitive financial information and professional responsibility.

One of our most important design principles was therefore:

Automate the work, not the accountability.

Andrew AI is designed so that AI can perform repetitive operational tasks while accountants retain visibility and control.

The platform includes audit logging and review workflows so activity can be traced back to the person or system responsible.

This creates a model where AI increases the capacity of accountants without removing human oversight.

Challenges we ran into

Building for real accounting workflows

The hardest part was not connecting an LLM to a user interface.

The challenge was understanding how an accountancy practice actually operates.

Documents arrive through different channels. Clients forget information. Transactions can be ambiguous. Accountants need working papers. Some actions can be automated, while others require professional judgement.

Andrew AI therefore had to be designed around real operational workflows rather than a generic AI chatbot experience.

Making AI trustworthy

Financial workflows require accuracy, traceability and accountability.

We needed to create safeguards around automated processing, including human review, audit trails, document status tracking and operational controls.

Connecting AI to existing tools

We also learned that transformation does not necessarily mean replacing everything.

Accountants already understand tools such as spreadsheets and email.

Andrew AI uses AI to connect and automate these familiar environments rather than forcing accountants into an entirely unfamiliar workflow.

Accomplishments that we're proud of

We are particularly proud that Andrew AI has moved beyond an idea or prototype chatbot.

We built a working system around a real accountancy practice and real accounting workflows.

During development we were able to:

  • Transform a highly manual accountancy workflow into an AI-assisted digital operation
  • Use Gemini for real financial document understanding
  • Build automated document intake and classification
  • Create structured client working papers
  • Build missing-document detection and client reminder workflows
  • Introduce human review and operational controls
  • Build detailed audit logging for team accountability
  • Create dashboards showing the operational status of clients and documents
  • Demonstrate how a small professional services business can become AI-native without removing the professionals

Most importantly, Andrew AI demonstrates that advanced AI infrastructure does not only have to benefit large enterprises.

It can give a small, established accountancy practice access to capabilities that previously would have required a much larger operations team.

What we learned

The biggest lesson was that the opportunity for AI in professional services is not simply replacing experts.

It is removing the enormous amount of repetitive work surrounding those experts.

Accountants should spend their time understanding businesses, identifying problems, providing advice and making professional judgements.

They should not spend hours manually moving information between emails, PDFs and spreadsheets.

We also learned that AI becomes substantially more valuable when it is embedded into an end-to-end workflow rather than offered as a standalone chatbot.

And perhaps most importantly:

Traditional businesses can become AI-native much faster than most people expect.

What's next for Andrew AI

Our next goal is to take Andrew AI from one accountancy practice to many.

Future development includes:

  • Advanced AI bookkeeping workflows
  • Bank reconciliation
  • Journal automation
  • Trial balance generation
  • Draft final accounts
  • Cash-flow forecasting
  • AI-assisted tax planning
  • Automated client onboarding
  • Multi-firm management
  • White-label deployments for accountancy practices

The architecture could eventually extend beyond accounting into other professional services where highly trained people still spend significant amounts of time on repetitive administrative work.

Our vision

There are thousands of small accountancy practices facing the same challenge as Andrew & Co.

They have decades of professional expertise and trusted client relationships, but often lack the technology teams and resources available to the world's largest accounting firms.

Andrew AI is designed to change that.

Our vision is to give small professional firms an AI operating layer that dramatically increases the amount of work each professional can handle while preserving human accountability.

We started with one 30-year-old accountancy practice in London.

Our ambition is to create a blueprint for how millions of traditional small businesses can become AI-native with Gemini.

Built With

Share this project:

Updates

Submission history