Case Studies Needed + ChatGPT just dropped a lot.

I'm looking for two agency founders who want to go through Agency OS Core — my 12-week coaching program — at 50% off the regular investment.

In exchange: you complete the program, you share your results, at the end you record a video testimonial about your experience, and we can use your program results as marketing and social proof. That's the whole deal. No catch, no extra asks.

We recently announced this program, which is built for agencies doing $8K–$50K/month who know their systems need work and are ready to actually fix them. If you've been on the fence about Agency OS Core, this is the most direct path in.

Two spots. First come, first served.

Reply to this email with the word "case study" and I'll send you the details.

OpenAI had a big week.

GPT-5.5 dropped. Images 2.0 launched. The model lineup got a full overhaul. And if you follow any AI content creators, your feed probably looked like a press release factory for about 72 hours straight.

Most of the coverage missed the point.

Listing features isn't the same as knowing what to do with them. So this week, I'm cutting through the noise. Here's what actually changed, what it means if you run an agency, and — because I'd be doing you a disservice otherwise — where Claude still beats it.

No hype. Just the operational read.

Four updates worth your attention. Everything else is noise.

01  GPT-5.5 — The Agentic Leap

This is the headline update and it's legitimately significant. GPT-5.5 isn't just a smarter chatbot — it's built to run multi-step tasks from start to finish without you managing every move.

OpenAI's own teams are using it to automate Slack request routing, process thousands of tax documents, and generate weekly business reports that previously took 5–10 hours a week to produce. Those aren't demos. Those are internal workflows.

For agency founders: the most relevant use case is document-heavy, multi-step work — research projects, report generation, briefing documents, competitive analyses. If you've got a task that requires pulling information from multiple sources and assembling it into a coherent output, GPT-5.5 is now meaningfully better at completing that loop without you babysitting it.

Available on Plus, Pro, Business, and Enterprise plans. GPT-5.5 Pro (the maximum capability version) is on Pro, Business, and Enterprise.

02  ChatGPT Images 2.0 — Text Finally Works

The previous version of ChatGPT's image generation had one persistent problem: text in images looked like a ransom note. Images 2.0 fixes this. Text renders accurately, aspect ratios are broader, and there's a "thinking" mode that reasons through complex visual requests before generating.

For agency founders: if you've been using Midjourney or other tools purely because ChatGPT mangled any image with words in it, that workaround is over. Social graphics, presentation covers, newsletter headers with text callouts — these are now viable inside ChatGPT. The thinking mode is paywalled (paid plans only) but the standard version is free.

One note here is that I still love Claude for generating graphics as well. We use it to create our newsletter graphics and social carousels. And NotebookLM is an awesome tool for infographics. Each tool has it’s place.

03  File Library — Your Work Stops Disappearing

Every file you upload or create in ChatGPT is now automatically saved to a Library. Searchable, browsable, reusable across conversations. Generated images go to a separate Images tab.

This sounds minor. It isn't. One of the biggest friction points with ChatGPT for ongoing work has been that everything lived inside individual conversations — upload a document in one chat, and it's gone the next time you open a new one. The Library changes that.

For agency founders: if you've been using ChatGPT for document work (proposals, briefs, reports), you can now build a library of agency assets — brief templates, style guides, scope documents — and reference them across sessions. It's a step toward ChatGPT functioning more like a persistent workspace.

04  Simplified Model Picker — Less Confusion, Finally

OpenAI killed the old GPT-5.1 models and overhauled the model picker. You now choose from three modes: Instant (fast, everyday tasks), Thinking (deeper reasoning, slower), and Pro (maximum capability).

The older model names are being retired. If you have any team members or workflows that reference specific model names like GPT-4o, those need updating — GPT-4o was fully retired at the start of April.

Here's my honest take on the AI landscape right now, as someone who uses both tools every day and installs them inside real agencies:

ChatGPT and Claude are not interchangeable. They're complementary.

The mistake most agency founders make is picking one tool and trying to use it for everything. The smarter move is knowing which tool wins in which situation — and routing work accordingly.

Use ChatGPT (GPT-5.5) when:

→       The task is agentic — multi-step, multi-source, needs to use tools and keep going

→       You need image generation with accurate text rendered in it

→       You're doing deep research across live web sources with a clear deliverable at the end

→       You want to build simple automations inside a familiar chat interface

Use Claude when:

→       You need sustained, high-quality writing that sounds like a human — not AI

→       You're building system prompts and agent architecture that needs to follow complex instructions reliably

→       You're working with long documents — contracts, transcripts, multi-page briefs — that need real comprehension, not summary

→       You want a consistent brand voice maintained across a long content workflow

→       You're building managed agents for your agency ops (the AGENT build spec we released last week is Claude-native for a reason)

The benchmark that matters for agency work isn't who scores highest on academic tests. It's who produces better outputs on the actual tasks your team does every day.

Run both tools on the same real task from your agency this week. The result will tell you more than any review article.

Pick one recurring content or research task in your agency that currently takes 2+ hours. Something with a clear output — a competitive brief, a monthly report, a new business deck, a client-facing summary.

Run it in GPT-5.5 (Thinking mode if you have a paid plan). Don't prompt-engineer it. Just describe the task as you'd explain it to a senior team member and let it run.

Then run the same task in Claude.

Compare the outputs on three dimensions: quality of the final deliverable, how much you had to correct or redirect it, and how long each actually took with your involvement included.

That comparison is your agency's AI routing guide. Build from there.

Once you've run the comparison above, use this Claude (or ChatGPT, but I still love Claude for strategy) prompt to turn your findings into a decision framework your whole team can use:

The output becomes a one-page AI routing doc you can share with your team. It replaces the "which AI should I use for this?" conversation — which, left unanswered, means everyone defaults to whichever tool they're most familiar with.

🛠️ TOOL OF THE WEEK ChatGPT Plus — $20/month

Given this week's topic, it would be odd not to feature it.

If you're using ChatGPT on the free plan, you're using a different product than the one I described above. GPT-5.5 Thinking, the File Library in full, and Images 2.0 thinking mode are all paywalled. The free plan gets the standard models with usage limits.

For agency use, the Plus plan at $20/month is the minimum viable tier. It gets you GPT-5.5 (standard), the File Library, and Images 2.0 standard. If you're using it for billable client work, that's an easy ROI calculation.

Business plan ($30/user/month) adds team workspaces, shared conversation history, and admin controls — worth it once you have 3+ people using it regularly.

One honest caveat: ChatGPT Plus is worth it as a second tool. If you're choosing between ChatGPT Plus and Claude Pro as your only AI subscription, that's a different question — and the answer depends on what your agency actually does. If you write a lot, Claude. If you do a lot of research and multi-step document work, ChatGPT. If you do both, you'll want both (my suggestion).

📊 BY THE NUMBERS

A number from OpenAI's own internal data that I keep thinking about:

More than 85% of OpenAI's own employees use Codex (their agentic coding tool) every week — across finance, marketing, communications, and product, not just engineering.

That's not a marketing stat. That's what it looks like when a company actually deploys AI inside real workflows rather than talking about it. Their finance team reviewed 71,637 pages of tax documents in a fraction of the normal time. Their comms team automated low-risk speaker request routing entirely.

The agencies that will look back on 2025–2026 as a turning point are the ones building these workflows now, while most of their competitors are still debating which AI tool to start with.

📣 A QUICK NOTE

The Claude Managed Agents Build Spec is still available if you missed it last week — full system prompts for three agents that run your agency ops 24/7, copy-paste ready, with deployment instructions and a cost breakdown. A 24/7 agent stack runs ~$58/month. A contractor doing the same work runs $800–2,000.

[Download the Build Spec here →]

And if you're ready to go deeper than a free resource — the Agency OS Lab community has the install files, SOP templates, and workflow builds that sit behind everything I write about in this newsletter. [Join the waitlist→]

See you next week.

Work smart. Enjoy life harder.

Founder, Agency Owner Lab

When you're ready, here's how we can work together:

→ Agency OS Lab — in the earlier stages of your business? DIY + community. SOPs, AI installs, tool stacks. [Join the waitlist here]

→ Agency OS Core — We design how your agency should run. 8-12 weeks. Like hiring a COO to architect your business. [Apply here]

→ Agency OS Accelerate — 12-month program. We stay with you as your agency evolves and scales. Like keeping a COO involved as your business grows. [Apply here]

→ The Boardroom — mastermind for founders at or approaching $1-3M/month. Life + business alignment with exclusive partner offers/resources. [Apply here]

→ Implementation Sprints — done-for-you systems builds, Standalone or paired with Agency OS. [Book a Systems Audit]

→ Strategic Reserve — on-demand strategic 1:1 advisory for solo founders and small teams. [Apply Here]