AI throughout. Built for the firms
that already run on it.

AI is not a module in The Link, and it is not a roadmap. It is already in the places your team works — natural-language practice analytics, per-client dossiers, voice-to-email, bank statement extraction, AI-drafted campaigns. Below are 15 of them, each in production use today.

Build your deck Book a walkthrough

Tier 1

Insight & Intelligence

The AI that answers questions about your practice, your clients and the work in front of you.

Practice Intelligence

A natural-language CFO-in-a-box. Ask agentic questions about profitability, capacity, VAT alignment, billing or chronology and Claude pulls the answer from your live practice data using ten-plus tools.

Where: Reports → Practice Intelligence. Pinned queries, scheduled runs, PDF export.

Client Intelligence

A nightly cron builds a living per-client dossier so the AI Advisor opens with full context. No cold start — instant answers about any client, any project, any deadline.

Where: Client record → AI Advisor. Refreshes overnight via the Anthropic Files API.

Client document analysis

On-demand analysis of OneDrive documents using firm-specific prompts. Read, classify and summarise client paperwork without leaving the client record.

Where: Client record → Document analysis.

Ticket Intelligence

Multi-turn AI with full ticket context — email, documents, bookkeeping and projects all in one conversation. Prompt caching keeps multi-turn costs roughly seventy-eight percent lower.

Where: Inbox → Unified Advisor on every ticket.

Email opportunity & sentiment

Every inbound email is classified for opportunity, urgency, sentiment and suggested replies — so revenue moments and risks are surfaced rather than buried.

Where: Inbox and ticket triage pipeline.

Tier 2

Voice & Transcription

Talk to your firm. Record a call. The transcript files itself.

Voice to email

Record a voice note. Whisper transcribes it. Claude formats it into a clean draft email or note ready to send. Real and used daily.

Where: Anywhere a draft can be created — inbox, client record, project.

Live captions & call transcripts

Live captions during video calls plus a permanent transcript auto-attached to the thread. Call ends, transcript is filed.

Where: Video calls and the RingCentral telephony integration.

Tier 3

Document & Data Extraction

AI that reads what your team would otherwise type by hand.

Bank statement AI extraction

Two-pass Claude extraction with bank-format caching. Repeat statements from the same bank cost roughly seventy-five percent less and come back faster.

Where: Bookkeeping → bank statement upload.

Tier 4

Generation

Drafts, sequences and pages — written for you, in your firm's voice.

AI email drafting & refinement

Generate or refine a draft email from a short prompt. The firm's house tone and system prompt are honoured.

Where: Email composer in the inbox.

AI campaign content

Drafts the subject line, the body and a full landing page for any campaign — context-aware to the campaign type and the segment.

Where: Campaigns → New campaign → Generate with AI.

AI reminder sequences

Generates the reminder sequences for PAR, CAR, onboarding and professional clearance — calibrated to the situation rather than copy-pasted templates.

Where: Reminders engine across onboarding and clearance flows.

Page builder AI

Conversational marketing page generation with URL-crawling. Describe the page you want, point at a reference URL, and the system builds it.

Where: Marketing pages module.

Text to structured note

Free-form text becomes a clean, bulleted note with the right structure for the client record.

Where: Notes panel on any client or project.

Tier 5

Workflow Intelligence

AI that helps the work land in the right place at the right time.

Workflow scheduling intelligence

Where workflow templates allow it, AI suggests start dates from deadline, complexity and team capacity — so jobs land on the calendar at the right time.

Where: Project scheduling on supported workflow templates.

AI-suggested tasks

Detects tasks that need to exist from project activity — so the team is reminded of work the system can see, not just what someone remembers to add.

Where: Project activity pipeline.

What we don't do with your data

Your data is never used to train any model — yours, ours or our providers'.
Anthropic and OpenAI are operated under enterprise terms with zero retention for training.
Prompt caching keeps your context short-lived in provider memory — never persisted to model weights.

The full data and access detail lives on the security page.

Models we use

Anthropic Claude powers the intelligence, extraction and generation features, across three tiers — a fast one for routing and short answers, a default one for real work, and a larger one held in reserve for the hard cases. OpenAI GPT handles email classification, document analysis and free-form drafting, and Whisper transcribes voice notes and live calls. We pick the right model for each job, not the cheapest one — and we move to newer ones as they ship, which is why this page names the families rather than the version numbers.

See AI in the context of your firm.

Build a personalised deck around your priorities — every outcome shows you which AI features earn their keep against the work you actually do.