With AI, what should we do in-house and what should outside counsel do?

When it comes up

During budget planning, panel reviews, when AI makes a category of work much faster, or when the business asks for faster turnaround.

Your options

Bring routine work in-house with AI

Fits when: High-volume, playbook-driven work: NDAs, standard contracts, first-pass research, regulatory digests.

Watch out: Needs tools, playbooks and capacity; don't overload a small team.

Keep with outside counsel on new AI-aware pricing

Fits when: Specialist, high-stakes or surge work, where firms use AI and share efficiency through fixed or capped fees.

Watch out: Firms may not pass on savings unless you ask and measure.

Alternative legal service providers with AI

Fits when: Large-volume review, contract migration and data extraction projects.

Watch out: Check quality controls, confidentiality and supervision duties.

Hybrid on a shared platform

Fits when: Ongoing matters where in-house and firm teams work together on the same AI-assisted analysis.

Watch out: Agree data ownership, access and confidentiality on the shared platform.

How to decide

Map matter types by volume, risk and specialism.
Ask what AI changes for each: speed, cost, required expertise.
Decide by tier: routine → in-house with AI or ALSP; specialist or high-stakes → outside counsel on AI-aware pricing.
Update guidelines for AI disclosure, confidentiality, verification and billing.
Measure spend and quality by matter type each quarter.

Trade-offs

In-sourcing saves fees but needs capacity, tools and supervision; outsourcing keeps flexibility and expertise but only saves money if pricing changes. Most departments end up hybrid.

Real examples

Microsoft's Corporate, External, and Legal Affairs organization, about 2,000 lawyers and compliance professionals, will use the Harvey platform for legal and compliance work, according to a July 2026 LawSites report. Source: LawSites (LawNext) — Microsoft's own legal department will use Harvey (23 July 2026) ↗
ABA Formal Opinion 512 (July 2024) says lawyers may not bill for time not actually spent and generally may not charge clients for time spent learning a general-use AI tool. Source: ABA Standing Committee on Ethics and Professional Responsibility — Formal Opinion 512: Generative Artificial Intelligence Tools (29 July 2024) ↗ · National Law Review — American Bar Association issues formal opinion on use of generative AI tools ↗
In J.G. v. New York City Department of Education (S.D.N.Y. 2024), Judge Paul Engelmayer cut a firm's fee request roughly in half and called its reliance on ChatGPT to justify billing rates 'utterly and unusually unpersuasive'. Source: Law360 — Law firm scolded for 'misbegotten' ChatGPT use in fees bid (J.G. v. New York City Department of Education, S.D.N.Y. 2024) ↗

What to check first

What are your top matter types by spend, and which are routine?
How do your panel firms use AI, and how does it show up in bills?
Do you have the capacity and playbooks to in-source routine work?
Do your outside counsel guidelines cover AI?

Learn and use

Sources

Last reviewed 2026-10-01.