Our Take
Sherrod's distinction between exposing low-value work versus exposing low-value lawyers is the actual insight—but her claim that contract drafting will be mostly automated in a decade rests on faith, not deployed evidence.
Why it matters
In-house legal teams are making hiring and training bets right now. How Intel's approach to AI adoption succeeds or fails will shape whether other large enterprises follow the same playbook or pull back.
Do this week
Legal operations: Audit your department's time spend on repetitive tasks (contract drafting, FAQs, summarization) before the next budget cycle, because Sherrod's team is already retraining staff toward higher-value work—don't get caught without a plan.
Intel's Legal Team Is Betting on Selective Automation
Joy Sherrod, Director of Discovery and Associate General Counsel at Intel, gave an interview ahead of the Legal Innovators California conference in June. Her core argument: in-house legal departments should hire for legal expertise first, tech skills second, but train everyone to be competent with AI tools immediately.
Sherrod disagrees with the framing that in-house lawyers need to become technologists. She argues that AI's highest-value use in legal is research, summarization, and data analysis—work that benefits junior attorneys and law firm teams more than senior in-house counsel. By contrast, in-house attorneys add value through specialized legal judgment across multiple domains.
At Intel, the approach is explicit: recruit the best lawyers, then train them to incorporate AI into their work. Sherrod says they have a "robust training program" to get all attorneys comfortable with AI, though no specifics on adoption rates or current coverage are disclosed.
On the future of legal work itself, Sherrod predicts contract drafting and development will become largely automated within ten years, with only the largest or most complex agreements receiving heavy attorney negotiation. Patent analysis will speed up, and acquisition of patent portfolios will become more targeted. Again: no benchmarks or deployed examples are cited.
The Real Risk Is Misalignment, Not Replacement
Sherrod reframes the "AI will replace lawyers" debate as misleading. The actual exposure is that AI will expose work that isn't adding value—and whether individual attorneys can shift to higher-value tasks, not whether they become obsolete. This is pragmatic framing, but it also requires execution discipline.
She identifies a critical failure mode in tech adoption: teams introduce tools to solve problems that don't exist. Adoption stalls when tools don't address genuine pain points, and when stakeholder buy-in (especially executive) is missing. Intel is flagging this explicitly—a signal that internal resistance is real even at large enterprises with legal tech budgets.
On the regulatory front, Sherrod expects divergence, not harmonization. The EU and UK will tighten AI governance around privacy and ethics. The US will remain permissive. Multinational corporations will need regional expertise, not global templates. This has cost and staffing implications that will outlast any single tool deployment.
What Legal Teams Should Do Now
Map your current work by time spend and business value. Sherrod's model assumes you can cleanly separate repetitive, low-value work from strategic advice. If you can't, your AI adoption will fail internally because attorneys won't see the benefit of retraining.
Get executive buy-in before deploying any new tool. Sherrod doesn't say how Intel secured this, but the fact that she emphasizes it suggests internal friction is common.
Build regional compliance protocols if you operate across borders. One global AI governance policy will not survive contact with EU, UK, or emerging-market regulators. Start asking your General Counsel about regional legal AI governance, not just data residency.
Finally: Sherrod's claim that contract drafting will be "largely automated" in ten years is not based on published benchmarks or customer wins she's disclosed. It is a prediction, not evidence. Don't fund automation projects on prediction alone. Require pilots with measurable time savings first.