Why Generic AI Tools Don't Work for Your Firm — And What Purpose-Built Tools Are Getting Right

Published: March 14, 2026 | By: The Crossing Report | 6 min read

---

Summary

If you've been hesitant to adopt AI because your firm's compliance obligations make general tools feel risky, you're not wrong. General-purpose AI tools weren't built for the ethical walls, conflicts clearance, and privilege protection that professional services firms require. That gap is real — and the industry is now building to close it. Intapp Celeste, launched February 2026, is the clearest signal yet that enterprise-grade compliance AI has arrived for law, accounting, and consulting firms. Here's what that means for your firm right now.

---

The Problem That Generic AI Can't Solve

Every professional services firm owner who has tried to deploy AI into client work has hit the same wall sooner or later. The tool works fine in isolation — summarizes documents quickly, drafts client communications competently, pulls research accurately. And then one question stops everything:

What happens to client data when it goes into this tool?

For a firm with professional confidentiality obligations — whether that's attorney-client privilege, CPA confidentiality rules, or consultant confidentiality agreements — the answer from general-purpose AI tools is deeply unsatisfying. Your client's financials, your contract language, your litigation strategy: when you send that into ChatGPT or a standalone Claude session, you're placing it on infrastructure with no understanding of your specific client relationships and no mechanism to prevent it from influencing responses about other clients.

That's not a theoretical concern. It's the reason your state bar has been issuing guidance all year, and why ABA Formal Opinion 512 now requires lawyers to maintain reasonable understanding of AI capabilities before using AI in client work. The compliance concern is real. Firms that dismissed it are now running to catch up.

But here's what's changing in 2026: the industry has heard this concern, and it's responding.

---

What Purpose-Built Professional Services AI Actually Does Differently

Three features separate professional services AI from general-purpose tools:

Ethical walls. Also called information barriers. At a firm managing dozens of client matters simultaneously, certain people — and certain tools — must not have access to information from competing or conflicted matters. A general AI tool has no concept of client-level access control. Purpose-built tools are architected from the ground up to partition client data so that a query about Client A cannot surface information from Client B's files.

Privilege protection. Attorney-client privilege (and its equivalents in accounting and consulting) can be waived by inadvertent disclosure. When you feed privileged communications into an AI tool that logs, processes, or stores that data on shared infrastructure, the privilege analysis becomes complicated quickly. Purpose-built tools are designed with privilege protection as a first-class requirement — not as an afterthought policy.

Conflicts clearance. Before a law firm accepts a new client or matter, it screens for conflicts against existing clients and relationships. This process is fundamental to professional responsibility and to firm reputation. Purpose-built AI can be integrated directly into the conflicts clearance workflow — not as a separate check, but as part of how the AI operates from day one.

These aren't features you can add to ChatGPT with a policy memo and a disclaimer. They require a fundamentally different architecture.

---

The Enterprise Signal: Intapp Celeste

In February 2026, Intapp announced Celeste — an agentic AI platform purpose-built for law, accounting, consulting, and investment management firms. Celeste uses prebuilt or customizable AI agents to automate four workflows:

• Client intake — screening, onboarding documentation, conflicts clearance
• Matter management — automating the administrative overhead of running active client engagements
• Business development — AI-assisted client relationship intelligence and pipeline management
• Matter delivery — coordinating the handoffs and document workflows inside an active engagement

All of it runs within the firm's existing compliance framework. Ethical walls, privilege protection, and conflicts clearance aren't features you toggle on — they're structural. Built on Anthropic's Claude and Harvey, with Intapp's professional services data architecture underneath.

Intapp is not a startup. It serves approximately 2,400 of the world's leading professional and financial services firms. When Intapp ships an agentic AI platform, it's because its clients — firm administrators, managing partners, COOs at mid-to-large firms — demanded it.

The signal here isn't "use Intapp Celeste this week." It's more fundamental than that.

The signal is: the enterprise tier of professional services just validated the entire category. When Intapp builds this and Gartner names it to its 2026 Agentic AI Innovation list alongside Salesforce and Microsoft, the remaining argument that "AI doesn't work for professional services compliance" evaporates. It works. The question is only when the right-sized version reaches your firm.

---

What This Means for a 10-Person Firm Today

Intapp Celeste is in limited release and targets firms with existing Intapp infrastructure. It is not a tool a 10-person firm buys off the shelf this week. That's fine. This article isn't telling you to buy it.

Here's what this means for you:

1. Your compliance hesitation is valid — and solvable. If the reason you haven't deployed AI in client-facing workflows is that you couldn't figure out how to make it work within your confidentiality and privilege obligations, that concern is correct. But the solution isn't to wait indefinitely. It's to use tools that already have compliance architecture built in at your scale.

2. Purpose-built tools exist today at the small-firm level. You don't need Celeste to get compliance-aware AI. For law firms:

• August — legal AI built for small firms. Free trial, no sales call, no seat minimums. Designed specifically for contract review and NDA triage with law firm confidentiality in mind.
• CoCounsel from Thomson Reuters — $225/user/month, no seat minimums. Backed by Westlaw. Built for practitioners who need outputs they can stand behind.
• Clio Manage AI — AI embedded directly in Clio's practice management platform, within your existing matter and client structure. Your data stays in Clio's environment.

For accounting firms:

• Thomson Reuters Checkpoint Edge with CoCounsel — the compliance-aware research and drafting stack for accounting professionals.
• Black Ore Tax Autopilot — built for accounting firm workflows, not general-purpose drafting.

3. General AI tools have their place — it's just not in your client files. ChatGPT and standalone Claude are appropriate for internal operations that don't involve client-specific data: drafting firm communications, researching industry trends, building internal templates and standard processes. The compliance boundary is about client data, not about the tools themselves.

4. Celeste tells you where the market is going. In 12–18 months, the compliance-aware AI architecture that Intapp is building for large firms will exist in products priced for your firm. The firms that have already built AI-assisted workflows into their operations — even with the accessible tools available today — will be positioned to layer on the more powerful infrastructure when it arrives. The firms still waiting will be starting from scratch.

---

Your Action This Week

Law firm: Sign up for August's free trial (no sales call required). Run three of your most recent NDAs through it. The compliance architecture is built for law firm confidentiality. Your clients' data stays in your workflow — not in a general-purpose consumer AI session.

Accounting firm: Log into Thomson Reuters Checkpoint Edge and check whether the CoCounsel integration is active on your account. If it is, run one client's most recent research question through it. Note the citation chain. That's the difference between authoritative AI and general-purpose AI.

Consulting or staffing firm: Map where client-specific data currently touches your AI workflow — even informally. If someone on your team is pasting client materials into ChatGPT for a summary, you have a compliance gap that doesn't require new tools to close. A one-page internal policy naming what's acceptable (firm-general content) versus off-limits (client-specific data) is the immediate fix. That policy costs you 45 minutes and reduces your exposure significantly.

The purpose-built AI market for professional services is arriving. The compliance concerns that held you back are being engineered away. The action now is to get ahead — not by buying enterprise tools you're not ready for, but by using today's accessible tools correctly while the infrastructure catches up.

---

Related Reading:

• Thomson Reuters CoCounsel Claims "Human-Level" Work — What It Actually Means for Small Firms
• The Best AI Tools for Small Accounting and Law Firms (2026)
• The Legal Tech Shakeout Has Begun — How to Know If Your AI Tool Will Survive
• Smarsh's AI Agent Cuts Compliance Alert Volume by 60% — What Smaller Financial and Staffing Firms Need to Know
• Best AI Tools for Small Accounting and Law Firms (2026)

---

The Crossing Report helps professional services firm owners make the crossing from the old way of doing business to the new one. Subscribe free →