Dive Deeper
Transcript
Episode Overview
Ask Diane Honda how in-house legal teams should adopt AI and her answer starts with a default: yes. At Redis, where Honda is Chief Administrative Officer, legal approves any AI tool a business function wants to try unless a lightweight review of data privacy and confidentiality turns up a specific reason to block it.
Each function at Redis now runs at least one AI tool, and some run several, because legal cleared the path instead of guarding it. In this episode of CZ and Friends, Honda walks Cecilia Ziniti through the review process behind that record, the leadership philosophy that powers it, and why she believes in-house counsel will write the AI rules that courts and legislatures have yet to draft.
About Diane Honda
Diane Honda is the Chief Administrative Officer at Redis, the in-memory database company whose platform powers the short-term memory layer of agentic AI applications. She oversees legal, compliance, information security, IT, and HR.
Before Redis, Honda was General Counsel at Barracuda Networks, where she helped take the company public in 2013 and led through M&A and private equity transactions.
Her career began in engineering: a software engineering degree with a double major in industrial management from Carnegie Mellon, nearly 12 years at Hewlett-Packard rotating through software, channel marketing, legal, and finance, and a JD earned in evening classes while working full time. She serves on the boards of the Hillman Group and Lucidworks.
Key Takeaways
Default to yes on AI adoption. Redis approves AI tools unless a fast privacy and confidentiality review finds a specific reason to block them, and each approval speeds up the next one.
A quick review process beats a perfect framework. Honda's team published usage guardrails and set response speed as the goal instead of waiting on a full AI policy.
Routine legal work is moving in-house permanently. Research, benchmarking, first-pass drafting, and filings that once justified associate billing rates now take minutes with legal AI, so outside counsel gets reserved for deep, partner-level expertise.
Human-assisted AI is the low-risk phase. An attorney still reviews each redline today, and Honda expects trust in fully automated output to build the way trust in calculators did.
In-house counsel will write the AI rules that do not exist yet. Courts and legislatures need lawyers who understand the technology, and Honda sees that as the role's next evolution.
How Should In-House Legal Decide Which AI Tools to Approve?
At Redis, legal approves any AI tool a business function wants to try, unless a fast review for data privacy and confidentiality turns up a specific reason to block it. The default is yes, so legal runs a quick clearance instead of standing as a gate.
That posture is why every function at Redis now runs at least one AI tool, and each approval makes the next one faster.
Honda described the review that clears the path:
"It's almost like if we can't find a reason not to let them try it, then we're going to let them try it. It's a pretty much approved-by-default philosophy."
The result is broad, fast coverage. Engineering runs AI code-generation tools, the team uses AI note-takers and project managers, and several AI-based legal tools are already in place.
Do You Need an AI Policy Before Adopting AI Tools?
No. Honda's team published usage guardrails and made response speed the goal, so teams could start moving before a full AI policy existed.
Guardrails plus a quick answer change behavior, because people surface new tools instead of hiding them.
Honda explained why speed beats a perfect framework:
"It's created this fast approval review process that has allowed us to leverage the tools more broadly, more quickly."
When a colleague brings in a tool that is already cleared, legal can wave it through on the spot. The policy grows out of real use.
How Does AI Change the Law Firm Associate Model?
Routine, entry-level work moves in-house. Research, benchmarking charts, first-pass drafting, and simple filings that once justified associate billing rates now take minutes with legal AI, so companies keep that work internal.
Outside counsel gets reserved for deep, partner-level expertise, and Honda sees a tiered model forming around that split.
Diane Honda, Chief Administrative Officer at Redis, put the shift this way:
"It's just going to gravitate more of that entry-level work to in-house counsel with AI tools, and you're really only going to want to pay for the deep partner-level expertise of a lawyer in an outside firm."
She names the open question too, of where future expertise comes from as less training happens inside firms.
Who Is Responsible When AI Generates Legal Work Product?
Today, the human stays accountable. Most of what AI produces in-house, a redlined document or a code fix, is still human-assisted, so a person reviews the output and owns the risk.
Honda expects trust in fully automated output to build gradually, the way trust in calculators and software did, through testing and QA until teams get comfortable.
She explained why the human stays accountable:
"We haven't pulled the human out of the AI experience yet, and so you still have that human controlling the risk and managing that."
That human-assisted phase is what makes the default-to-yes posture low risk today.
Who Writes the Rules for AI-Generated Evidence?
In-house counsel, before the courts catch up. A document produced purely by AI has no human author to lay an evidentiary foundation or testify to its veracity, so the old rules of evidence do not fit cleanly.
Honda argues in-house teams should write the internal rules now and help educate judges while the law catches up.
She made the case for leading now:
"We in-house are going to have to make the rules that don't exist on the outside yet... I think it's a great opportunity for us to lead and not wait."
In practice that means internal policies today, telling teams how they can use AI and bringing the edge cases to legal for guardrails.
What Mindset Makes a Default-to-Yes Approach Work?
A bias toward action. Honda's adoption philosophy runs on a principle she draws from Burn the Boats by Matt Higgins, which is to commit fully to the plan in front of you.
Applied to AI, that means saying yes and learning fast.
She described how she leads:
"I have a bias towards action... When someone gives me an opportunity, I generally don't say no, because I'm always wanting to take things to the next level."
For a legal team weighing AI, the move is to say yes first, with guardrails to keep it safe.
Honda's default-to-yes only works with a platform that clears the privacy and confidentiality bar on the first pass. Ready to run that review on your own team?
Recommended Reading
AI for General Counsel Operations: one operating layer for solo GCs through full departments, a fit for Honda's multi-function CAO seat and her end-to-end view across legal, compliance, IT, and HR.
Why Judgment Is the New Superpower for In-House Lawyers in the Age of AI: extends Honda's human-in-the-loop stance on who stays accountable when AI does the work.
Burn the Boats: Diane Honda on What Best GCs Know About AI Adoption: GC AI's own write-up of this conversation and Honda's default-to-yes framework.
