# Imagine — For Tax & Audit Firms *Your firm already has the tools. Now give them a brain.* A pitch document for principals of tax, audit, and accounting firms who already use **Microsoft 365 + Copilot**, **Sage / Tally / Zoho Books**, and **ChatGPT**. --- ## 1. The honest question > "If I already have Microsoft 365 + Copilot for collaboration, Sage for accounting, and ChatGPT for drafting — what does Imagine give me that they don't?" Fair question. The short answer: those three are **excellent tools**. But each one does its job inside its own silo, and none of them — individually or together — builds the one thing a tax & audit firm actually runs on: **a memory of your firm**. | Tool | Great for | What it's *not* | |---|---|---| | **Microsoft 365 + Copilot** | Documents, email, Teams chat, in-the-moment AI inside Word/Excel | A colleague who remembers what your firm decided three months ago, or what a junior promised a client in March. Copilot is a smart cursor, not a colleague. | | **Sage / Tally / Zoho Books** | Ledgers, GST returns, P&L, balance sheets, statutory output | A reasoning associate. Sage records numbers — it doesn't follow up on missing documents, draft client communications, or remember each client's quirks. | | **ChatGPT** | Drafting, brainstorming, summarising in the moment | Your firm's memory. A fresh tab every time — it doesn't know your clients, your tone, your engagement letters, your prices, or what you decided yesterday. No audit trail. No reporting line. No "boss". | --- ## 2. The missing layer: **Organisational Intelligence** Every tax & audit firm is built on knowledge that lives in seven places at once: partner brains, WhatsApp threads, Excel sheets, Tally notes, email threads, that one senior who's been here 8 years, and that one client folder nobody else opens. **Imagine pulls that scattered knowledge into a single, queryable, living brain** that every employee, every agent, and every client interaction can draw from. ``` [ MS 365 ] [ Sage / Tally ] [ Emails / WhatsApp ] [ Partner & staff knowledge ] \ | | / \ | | / ──── Imagine — Organisational Intelligence Layer ──── / | | \ / | | \ [ Drafted client reply ] [ Internal checklist ] [ Compliance flag ] [ Partner-ready brief ] ``` The output is delivered in your firm's voice, with each client's history attached, and a full audit trail. --- ## 3. The four properties — why this becomes permanent These four properties are what turn AI from a productivity tool into a permanent member of the firm. Only Imagine offers all four at once. ### Agents that don't resign When a senior associate leaves, eight years of client context goes with them. **Imagine agents stay.** The institutional memory is owned by the firm, not the person. ### Agents that don't forget Three-layer memory: today's chat, this week's digest, forever-knowledge. A 6-month-old agent has the muscle memory of a 6-month employee — without ever being re-briefed. ### Agents that don't favour No internal politics. No "I'll do my friend's file first". The agent answers everyone — junior staff, senior partner, top client, marginal client — with the same standard, the same speed, the same rules. ### Agents that don't sleep A client emails at 11:42 PM on a Sunday during year-end? The agent answers, drafts the reply for your approval, and queues the documents you'll need. You wake up to *done*. --- ## 4. The four ingredients (per agent) Every agent is built from four plain markdown files you can read and edit: 1. **Soul** — Identity, tone, behaviour rules, knowledge boundaries, escalation triggers. The agent's job description. 2. **Three-layer memory** — Short (today, raw) → Medium (this week, digested) → Long (forever, durable). Auto-compacts every night, and every Sunday. 3. **Team groups** — Agents live *inside* your firm's team chat (or Teams / WhatsApp / Telegram). Anyone can `@mention` them. 4. **Reports-to line** — Every agent names a human boss (a partner). Out-of-scope or low-confidence questions escalate *to that human, by name*. **No black boxes.** Read or edit any file, anytime. --- ## 5. Six agents most tax & audit firms hire in week one | `@handle` | Job | |---|---| | `@client-care` | Answers "what documents do you still need from me?", drafts polite follow-ups, knows each client's history and preferred tone. | | `@audit-buddy` | For any engagement, surfaces the SA-checklist, flags missing working papers, prevents the embarrassing "we forgot 3CD para 17" moment. | | `@compliance` | Reads every CBDT & CBIC notification, pushes a 3-bullet summary to `#firm-updates`, calls out which clients are affected. | | `@mentor` | Walks article-clerks through firm SOPs, answers their dumb questions kindly, flags real gaps in training material to a partner. | | `@brief` | Sunday night, drops a 7-bullet recap of open engagements, decisions taken, missed deadlines, and money still due. | | `@billing` | Tracks open invoices, drafts polite reminders, escalates anything above your threshold to a partner. | **Setup time per agent: ~3 minutes** — name, one-line job description, reports-to. Imagine drafts the agent's soul.md for your approval. --- ## 6. A Tuesday during October (peak audit season) - **07:55** — `@brief` drops the weekly summary in `#partners`. You read it over coffee instead of scrolling 200 emails. - **10:14** — `@audit-buddy` flags: *"Client X's 3CD draft is missing Para 17 (R&D capex) and a related-party note in Para 23."* The fresher didn't even ask — the agent volunteered it. - **12:30** — `@compliance` posts: *"New CBIC notification 12/2026 hits 3 retail clients. Drafted a 1-pager for each."* You glance, approve, send. - **15:40** — `@client-care` drafts a reply to a TDS-on-professional-fees query, pulling the client's exact history, your firm's stance, and last year's similar query. You edit one line. Send. - **22:10** — A panicked client messages: *"Got a 143(2) notice today, what do I do?"* `@client-care` acknowledges within 2 minutes, explains the next 7 days in plain English, books a slot on your calendar. You sleep. - **00:00** — Every agent goes to sleep for 10 minutes. Today's chat distils into tomorrow's medium-term memory. **You did nothing. The firm got smarter overnight.** --- ## 7. Side by side — where Imagine fits | | MS 365 + Copilot | Sage / Tally | ChatGPT | **+ Imagine** | |---|---|---|---|---| | Edits docs in real time | ✅ | ❌ | partial | ✅ | | Posts statutory ledger | ❌ | ✅ | ❌ | reads & references | | Remembers **your firm** across sessions | ❌ | ❌ | ❌ | ✅ 3-layer memory | | Remembers each **client** across years | ❌ | numbers only | ❌ | ✅ | | Named agents with a job description | ❌ | ❌ | ❌ | ✅ | | Lives inside team chat — `@mention` | Teams only | ❌ | ❌ | ✅ Teams, WhatsApp, Telegram | | Escalates to a **named human** | ❌ | ❌ | ❌ | ✅ reports-to line | | Audit trail of every reply | partial | within Sage | ❌ | ✅ every reply timestamped | | 24×7 client response | out-of-office only | ❌ | generic | firm-aware, your voice | | Survives a senior leaving | ❌ | numbers stay | ❌ | ✅ institutional memory | | Your data, on your server | MS cloud | Sage cloud | OpenAI cloud | ✅ markdown on disk | **Imagine doesn't replace MS, Sage or ChatGPT.** It connects them with a memory — so the firm's knowledge compounds instead of leaking. --- ## 8. Knowledge that compounds By month six, the firm's long-term memory typically holds: **By client** — communication tone they prefer · their typical delays & quirks · standing observations from prior years · what you've already explained to them · pricing arrangement & payment behaviour. **By practice area** — firm's stance on grey-area positions · standard reply templates per notice type · latest case law you've cited · internal QC checklists · "things that bit us last year" register. **By person** — who's strongest on which client · who carries which institutional knowledge · onboarding flow that's improved over time · open feedback & coaching notes · workload picture. When a senior leaves, you don't lose any of this. When you hire a junior, they walk into a firm with **memory**. --- ## 9. Governance — an audit firm cares about controls. So does Imagine. - **Approval before any agent speaks** — new agents are `pending` until a partner approves. - **Every agent reports to a partner** — out-of-scope questions escalate to that partner, by name. - **Full audit trail** — every reply, every escalation, every approval is timestamped and queryable. - **Editable memory** — don't like something the agent learned? Open the markdown file, edit, save. - **Your server, your data** — memory is plain markdown on disk. Export, back up, delete at any moment. - **Sleep windows & mute** — take an agent offline for a weekend or a sensitive client with one click. --- ## 10. The math | | | |---|---| | Partner time reclaimed | **8–12 hours / week** across the day-one bench | | Setup time per agent | **~3 minutes** | | Knowledge lost when a senior leaves | **0 hours** | The intangible: your **SOPs finally get written down**. Every soul.md and long-term memory file is, by accident, the cleanest piece of practice documentation your firm has ever had — exactly what every CA practice review keeps recommending you build. --- ## 11. Get started A 30-minute call. We'll set up your first two agents on a real client engagement of your choice, with you in the chat watching every reply. If by the end of week one you don't see the partner-time savings, walk away — no commitment, no cost. **Keep Microsoft. Keep Sage. Keep ChatGPT. Add the brain.** --- *imagine.fadafad.com — the organisational intelligence layer for professional services firms.*