AgentsAutonomy & tool use
Gilbert + Tobin CEO says ChatGPT Work agent assembled 120 partner review packs for about $62
Gilbert + Tobin's chief executive says an OpenAI agent built confidential annual-review packs for 120 partners in minutes for about $62. The Australian law firm reports 87% active use among staff with ChatGPT access after a rollout that set no usage targets and put governance first.
From the OpenAI News article of November 12, 2026 and cited sources: CEO Sam Nickless used the ChatGPT Work agent to update partner files from a spreadsheet, generate PDFs and bundle them. [1] [2] [3]
SOURCES[1] OpenAI: Gilbert + Tobin × Sam Nickless openai.com[2] StartupHub.ai startuphub.ai[3] Inside AI insideai.news
AT Pulse research deskPublished · Updated · 4 min read · AI-assisted analysisAI-assisted, not human-edited. Methodology Corrections Who owns this site
- What happened
- CEO Sam Nickless used the ChatGPT Work agent to update partner files from a spreadsheet, generate PDFs and bundle them. [1] [2] [3]
- Original source
- OpenAI News article, company source
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- When
- November 12, 2026 article published
- Covered
- Nov 12 edition
- Evidence
- 15 sources across 12 sites
Reach: not measured yet
CEO Sam Nickless used the ChatGPT Work agent to update partner files from a spreadsheet, generate PDFs and bundle them. He says the job took minutes and about $62, against what would otherwise have been several people working for a few days. The workflow is a structured batch document task, not legal reasoning. [1] [2] [3]
In June 2026, 87% of staff given ChatGPT access were active users. The firm says that is more than twice its usual uptake for new tools. It credits demos, challenges and visible use by leaders, including a custom GPT 'digital twin' of the CEO that his direct reports use to test ideas, rather than mandates. [4] [7]
The rollout was staged by risk. It began with operations teams, then moved to marketing, recruitment, finance, technology and selected practice areas. Client data was allowed only later, once Australian data residency, contractual protections and role-based admin controls were in place. Staff must review outputs, and people stay accountable for decisions. [3] [5] [10]
The figures come from OpenAI and the firm, and no baselines or error rates are published. If they hold, law firms may get early agent gains from structured operations work once data rules are settled, not from usage mandates.
Sources are gathered; writing the dossier now.
Executive brief
Gilbert + Tobin CEO Sam Nickless says he used OpenAI's ChatGPT Work agent to put together confidential annual-review packs for 120 partners. It took minutes, cost about $62, and he says the same work would otherwise have taken several people a few days (StartupHub.ai; Inside AI). The Australian firm also reports that 87% of staff with ChatGPT access were active users in June 2026, with no usage targets (OpenAI). OpenAI published the Q&A with Nickless on 12 November 2026. Its message is that the firm's results came from leadership by example and governance, not from cutting staff.
What changed and event timeline
Australian data residency arrives
OpenAI let ChatGPT Enterprise and Edu customers keep their data in Australia, among other regions, at no extra cost (;). The firm says this gave it the confidence to widen access.
87% of enabled seats active
The firm reports active use more than twice its usual rate for new tools. It set no mandates and relied on demos and challenges instead (;).
ChatGPT Work launches
OpenAI released an agent with Codex built in that runs multi-step tasks across apps and files. It is powered by GPT-5.6 and rolled out first to Pro, Enterprise and Edu users (;).
OpenAI customer story
The case study describes a rollout that started with operations teams, then expanded to marketing, recruitment, finance, technology and selected legal practice areas ().
11–12 Nov 2026: Nickless Q&A
OpenAI's interview covers the $62 partner-review workflow, a "digital twin" custom GPT and the firm's three adoption phases (;).
Capabilities and access
- Tools in use: ChatGPT Enterprise, ChatGPT Work, Codex and custom GPTs (StartupHub.ai).
- Model: ChatGPT Work ran on GPT-5.6 at launch (iGeeksBlog). The firm's coverage does not name the model behind its own workflows.
- Partner reviews: the agent updated files from a spreadsheet, generated PDFs and bundled the results.
Read the full section
- Tools in use: ChatGPT Enterprise, ChatGPT Work, Codex and custom GPTs (StartupHub.ai).
- Model: ChatGPT Work ran on GPT-5.6 at launch (iGeeksBlog). The firm's coverage does not name the model behind its own workflows.
- Partner reviews: the agent updated files from a spreadsheet, generated PDFs and bundled the results.
- Digital twin: a custom GPT built on Nickless's writing and feedback. His direct reports use it to test ideas before meetings. It does not make decisions or speak for him (Superpower Daily).
Technical analysis for researchers and developers
- Workflow shape: the partner-review job is a deterministic batch pipeline: structured spreadsheet in, templated documents and PDFs out, then bundling.
- Access controls: Australian data residency, contractual protections, role-based access and admin controls (Inside AI).
- Rollout by risk tier: client data was barred at first and allowed later under these protections.
Read the full section
- Workflow shape: the partner-review job is a deterministic batch pipeline: structured spreadsheet in, templated documents and PDFs out, then bundling. That is low-risk agentic work, not legal reasoning.
- Access controls: Australian data residency, contractual protections, role-based access and admin controls (Inside AI).
- Rollout by risk tier: client data was barred at first and allowed later under these protections.
- Not documented: no evaluation method, error rates, prompts or review steps are published, so the results cannot be reproduced.
Claims and evidence
- 87% active use: the figure covers only staff given access, not the whole firm.
- $62 for 120 partner reviews: also self-reported. No baseline staff-hour figure is published (StartupHub.ai).
- Reported time savings:
Read the full section
- 87% active use: the figure covers only staff given access, not the whole firm. It is a customer-reported figure published by OpenAI and has not been independently checked (Superpower Daily).
- $62 for 120 partner reviews: also self-reported. No baseline staff-hour figure is published (StartupHub.ai).
- Reported time savings:
- Recruitment research: about 4 hours down to 20 minutes.
- Conflict, KYC and AML checks: 3–8 hours down to 5 minutes.
- Audit reporting across 300 entities: about a day of manual work avoided (AIFrontWire).
- Independent corroboration: none was found. The secondary coverage restates OpenAI's materials.
Context and prior work
Read the full section
- Gilbert + Tobin was founded in 1988 and has more than 500 lawyers across Sydney, Melbourne and Perth. Nickless became CEO in 2022, succeeding Danny Gilbert (Wikipedia).
- Nickless had said earlier that by 2025–26 firms would need a good suite of AI tools or be left behind (LMHub).
- ChatGPT Work competes directly with Anthropic's Claude Cowork (Reworked).
- OpenAI's push into Australia also includes a A$7 billion data-centre partnership with NEXTDC (OpenAI).
Limitations, safety and contested findings
- All performance figures come from OpenAI and the firm itself.
- The governance rules require staff to review outputs and keep humans accountable.
- The documented wins are mostly in operations and back office (reviews, KYC, recruitment). There is little published evidence about client-facing legal advice.
Read the full section
- All performance figures come from OpenAI and the firm itself.
- The governance rules require staff to review outputs and keep humans accountable. AI is framed as help with tasks, not as an autonomous decision-maker (AIFrontWire).
- The documented wins are mostly in operations and back office (reviews, KYC, recruitment). There is little published evidence about client-facing legal advice.
- "Rethinking the business of law" is the headline's framing. The reviewed sources say nothing about changes to pricing or the billable hour.
Business and practitioner implications
- Lead from the top: visible use by the CEO, not mandates, is the firm's stated driver of adoption.
- Settle data rules first: data residency and contract terms made it possible to allow client data later.
- Start in operations: back-office workflows with structured inputs are the fastest early wins.
Read the full section
- Lead from the top: visible use by the CEO, not mandates, is the firm's stated driver of adoption.
- Settle data rules first: data residency and contract terms made it possible to allow client data later.
- Start in operations: back-office workflows with structured inputs are the fastest early wins.
- Workforce framing: Nickless says the value so far has come from freeing people for the more human parts of their work, not from cutting cost or headcount (Inside AI).
- Measure before you quote: buyers should establish their own baselines before relying on vendor-published savings.