AI Features Guide

The AI reads the document content and produces a concise executive summary covering the main purpose, key figures, dates, parties, and any notable clauses or findings. The summary is displayed in a formatted overlay panel.

The summary is generated on-demand and not stored permanently — click the button again to regenerate with a fresh call.

The AI analyses the document and returns a structured extract of key data fields: dates, amounts, account numbers, party names, tax figures, and any other domain-relevant fields it can identify. The extracted data is displayed in a formatted table ready for copying.

For financial documents, the extractor specifically looks for revenue totals, net profit/loss, tax liabilities, balance sheet line items, and invoice amounts. For legal documents, it extracts parties, dates, obligations, and key terms.

Many clients upload files with unhelpful names like "scan001.pdf" or "document (3).xlsx". The AI reads the document content and suggests a descriptive, consistent filename following your firm's naming convention. You can accept the suggestion with one click to rename the file instantly.

Suggested filenames typically follow the format: ClientName_DocumentType_Period_Date (e.g., "SmithLtd_AnnualAccounts_FY2024_20240901.pdf").

Type any natural-language question about the document — such as "What is the net profit for 2023?", "Who are the parties to this agreement?", or "What is the depreciation figure on page 4?" — and the AI will search the document content and answer your question directly, citing the relevant section or page.

The Q&A interface stays open so you can ask follow-up questions about the same document in a conversational flow. Each response includes a confidence indicator and source reference where possible.

The AI reviews the document for potential anomalies, inconsistencies, or items that warrant closer attention. For financial documents this includes unusual rounding, figures that don't reconcile between sections, missing required line items, or values significantly outside expected ranges.

Each flagged item is presented with a description of the anomaly and a suggested action (e.g., "Verify", "Clarify with client", "Check against prior year"). Flags are advisory only — human judgement remains the final authority.

Based on a document (such as a set of accounts or a tax return), the AI identifies related documents that are referenced but not yet uploaded to the vault. For example, if accounts reference a rental income schedule, the AI flags that the schedule has not been uploaded.

Results are displayed as a checklist of missing items you can send to the client as a document request with one click, pre-populating the request form with the identified missing items.

Before sending a message, use Tone Check to get AI feedback on whether your draft sounds professional, empathetic, and appropriate for a client-facing communication. The AI evaluates your text and returns a traffic-light rating:

• ✅ Professional — tone is clear, respectful, and appropriate
• ⚠️ Neutral — acceptable but could be improved
• ❌ Informal / Harsh — rewrite recommended

Along with the rating, the AI provides specific feedback and offers a rewritten version of your message if improvements are suggested.

The AI reads the entire thread history and drafts a professional, contextually appropriate reply. The draft takes into account your firm name, the staff member's name, the client's name, and the full conversation context to produce a reply that sounds like it came from your team.

The draft appears directly in the reply text box where you can edit, shorten, or expand it before sending. You are always in control — the draft is a starting point, not a final answer.

For threads with many messages, the AI produces a concise summary of the conversation covering: what was originally asked or raised, what was agreed or actioned, what items remain outstanding, and any key dates or deadlines mentioned in the thread.

The summary is formatted as an easy-to-read structured list, not just a paragraph, making it quick to share internally or attach to a file note.

Describe a piece of work in plain English — for example, "Prepare the year-end accounts for ABC Ltd including reconciliations, P&L, balance sheet, and CT600 filing" — and the AI will break this down into a structured list of individual tasks with suggested titles, descriptions, and priorities.

Review the generated task list in the preview panel. You can remove any tasks you don't need, edit titles or descriptions, and adjust priorities before clicking Create All to add them to the task board in one operation.

The AI reviews all your open and in-progress tasks — including their titles, descriptions, priorities, due dates, and associated clients — and produces a recommended execution order with a brief reasoning statement for each decision.

The output is a numbered prioritised list you can use to plan your day or week, delegate to a team member, or share in a team standup. Decisions consider deadline proximity, client importance indicators, and task interdependencies.

Enter the client's name, service types (e.g., "annual accounts, personal tax return, VAT"), and your firm name. The AI generates a full, professional engagement letter including: scope of services, fee arrangement placeholder, confidentiality provisions, limitation of liability clause, and signature blocks for both parties.

The generated letter is editable before use — review and amend as appropriate, then attach it directly to a signature request to send for e-signature. The letter is also accessible from the client profile page if you prefer to generate it outside the signature flow.

Based on the client's profile — including business type, industry, services engaged, and jurisdiction — the AI produces a tailored onboarding checklist covering everything from initial KYC documents and bank details through to setting up recurring billing and scheduling the first review meeting.

The checklist is interactive and can be assigned to a staff member, with items checked off as they are completed. It can be shared with the client so they can see what is still outstanding from their side.

Set a date range in the Audit Trail and click Analyse Period. The AI reads the audit log entries for that period and produces a narrative report covering: total event counts by type, most active users, peak activity periods, and any notable patterns worth highlighting for compliance purposes.

The output is formatted as a structured report suitable for inclusion in a compliance file or presenting to a manager/director. Event types include document uploads, downloads, logins, signature requests, and data modifications.

The AI scans recent audit log entries looking for anomalous patterns that may indicate security issues or policy violations. Patterns it looks for include: unusual login times (outside normal working hours), bulk document downloads by a single user, multiple failed login attempts, activity from unexpected IP addresses, and unusually high transaction or event volumes.

Each flagged item is presented with a risk level (High / Medium / Low) and a plain-English description of why it was flagged. Items can be marked as "Reviewed" or escalated for further investigation directly from the results panel.

The AI produces a comprehensive written status report for the selected client covering: active engagements and their status, document submission status and outstanding requests, task progress and overdue items, a summary of recent communications, outstanding invoices, and key upcoming dates or deadlines.

The report is formatted professionally and is suitable for printing and reviewing with the client in a meeting. It can also be accessed from the client profile's Reports tab for quick on-demand generation before any client interaction.

Type a question in plain English on any data-heavy page — such as "Show me all overdue tasks for Smith Accounting", "Which clients haven't uploaded any documents this month?", or "Find invoices over £5,000 that are more than 30 days overdue". The AI interprets your query and converts it into a structured data filter that is applied immediately to the current view.

The AI Search bar appears in the toolbar on the Documents, Tasks, Clients, and Audit pages. The interpreted filter is also shown so you can verify it matches your intent, and you can adjust it manually if needed.

The AI analyses your accounts receivable data and writes a plain-English narrative summary: total outstanding broken down by age bucket (current, 30-day, 60-day, 90+ day), identification of highest-risk debtors with specific amounts, suggested next actions for each overdue account (e.g., "Send reminder email", "Escalate to collections", "Consider payment plan"), and a brief cash flow outlook.

The narrative can be exported as a PDF or copied for inclusion in a management report, board pack, or cash flow review document.

Select one or more uncategorised or suspect transactions and click Suggest Categories. The AI reviews each transaction's description, amount, and date and recommends the most appropriate account category from your chart of accounts. Suggestions are displayed inline against each transaction.

You can review each suggestion and accept or decline individually, or use Bulk Accept All to accept all recommendations at once. The AI learns from the patterns in your existing categorised transactions to improve suggestions over time.

Select one or more invoices and click Explain Invoices. The AI generates a friendly, client-readable narrative explaining the charges on each invoice — particularly useful when invoice line items use internal codes or shorthand that may confuse clients (e.g., "CT600 prep" becomes "Preparation and submission of your company tax return for the period ending 31 March 2024").

The narrative is designed to be attached to the invoice email or included as a cover note, reducing client queries and improving payment rates by making invoices immediately understandable.