OpenAI just announced 'Harness Engineering,' their new internal system where AI agents, called Codex, are writing large-scale software.
The Story
They claim a beta product with a million lines of code was shipped in five months with no manually written source code, guided by a small team of engineers.
Why It Matters
You're not building large software applications, but the underlying lesson is vital. This isn't about AI replacing your team; it's about *delegating* complex, structured tasks. For your firm, it means using your internal knowledge – your procedures, your templates, your client history – as the 'single source of truth' that AI agents can then leverage to automate repetitive tasks, from drafting initial documents to managing client communication.
What To Do About It
My concrete advice: identify one repetitive, rule-based process in your firm – perhaps initial client intake forms, or drafting standard contracts. Spend the next month *structuring* your internal documentation around this process, making it explicit and machine-readable. Don't think about buying a fancy AI yet; focus on cleaning your house. This disciplined approach is how you prepare for effective AI delegation, saving you thousands later. Start this quarter.
Related Signals
Microsoft rolled out its Project Manager Agent worldwide in April 2026 as part of the Copilot 2026 Wave 1 release, alongside video meeting recaps and Community Agents in Teams.
10 Apr 2026
Salesforce overhauled Slackbot with 30+ new AI capabilities, turning it into an autonomous desktop agent that executes tasks via MCP integrations with Google Workspace, Microsoft 365, Notion, Workday, and 6,000+ apps.
4 Apr 2026
Salesforce embedded Agentforce AI agents into its Free, Starter, and Pro SMB Suites at no extra cost — capabilities previously locked behind $550/user enterprise plans now included in every SMB seat.
26 Mar 2026