Practice
Hallucination Is a Workflow Problem, Not a Model Problem
Retrieval, verification and citation checking turn unreliable outputs into dependable work product.
By Eleanor Whitcombe · October 1, 2026 · 7 min read

Every sanction order involving fabricated cases tells the same story: a lawyer trusted an output they never checked. The model failed, but the workflow failed first.
Modern legal AI tools ground answers in verified sources and link every citation. Yet even grounded tools can mischaracterize holdings. Verification remains a professional duty that cannot be delegated to software.
The most effective teams build verification into the process itself — automated citation checks, required source review, and sign-off steps that are logged.
Treat AI output as you would a memo from a talented but untested summer associate: useful, fast, and never filed without review.
§
Further reading

Practice
Why Small Firms May Win the AI Decade
Lean practices can adopt faster, price more creatively and market more personally than their larger rivals.
Daniel Ferreira · October 6, 2026 · 6 min

Practice
In-House Teams Are Now Leading AI Adoption
Legal departments, pressed to do more with less, are outpacing their outside counsel.
Priya Raman · September 11, 2026 · 7 min

Practice
The Knowledge Management Renaissance
Firms that organized their precedents are reaping rewards. Everyone else is scrambling.
Eleanor Whitcombe · August 22, 2026 · 7 min