Autonomous Growth Agent
Last week was a watershed moment in the world of internet products, especially for its impact on the enterprise.
OpenAI launched a variety of autonomous enterprise capabilities.
- A personal agent focused on work tasks called Dots that runs autonomously in the background
- An expanded Plugins ecosystem as an AI-native app store
- Sign In with ChatGPT for customers on eligible plans to use their existing AI allowance in participating apps (instead of wrangling your own compute; separate app fees may apply)
- Codex cloud computer in addition to running on your desktop computer
- AI accelerating OpenAI’s own research, bringing recursive self-improvement closer
- OpenAI reports an internal model resolved more than 100 long-standing open math problems
Taken together, I believe it enables autonomous organizations to exist for the first time. I should know, I’m building one.
So let me give you the steps to start driving revenue growth using these tools and create the first step toward your own autonomous organization that grows on autopilot.
Remember, all of AI is based on mathematics. What compounds is the progress: better tools help us build still better tools. My bet is that we are now in the era where the compounding is compounding, and 2027 is going to accelerate like nothing we’ve yet seen as a species.
So you need to be ready to start adapting in real-time to changes. Don’t think, just do :)
Steps to starting your own Autonomous Growth Agent.
- If you don’t already have the ChatGPT Pro tier (as a personal account) or Business Premium, sign up for it. Pro offers $100, $200, and $500 monthly tiers, with Pro 500 providing the highest usage allowance. Business Premium starts at $100 per user per month billed annually, or $125 billed monthly, with a minimum of two seats. Availability depends on rollout, plan eligibility, and admin settings.
- Download the ChatGPT / Codex desktop app.
- I’m using an Apple Mac for this workflow.
- Move all AI work to OpenAI.
- Anthropic takes the possibility of AI consciousness and model welfare seriously. My bet is that enterprise buyers will be uncomfortable with that approach for mission-critical systems and increasingly favor OpenAI.
- This is important because you want relevant, company-approved data being used as context, with the necessary permissions, for the agents to be more precise and personalized. This drives better outcomes, results, and requires less explaining over time.
- Open ChatGPT/Codex and set up your Dot.
- Your Dot should appear in the upper left side bar. Let’s start with one for simplicity’s sake.
- Every morning you should click ChatGPT > Check for Updates in the menu bar and update the app. The app is improving quickly, so install updates when they’re available.
- Give it your relevant, company-approved data and set up Plugins with the necessary permissions.
- Plugins are also in the left sidebar. Install all the ones for all the software you use, including Gmail, Calendar, Google Drive, Zoom, Stripe, Github, Dropbox, Hubspot, Salesforce, Quickbooks, Mercury, etc.
- As you can see Enterprise software is now a plugin inside ChatGPT. You can enable it for your enterprise software company as a new growth driver. That’s bonus #1.
- Alignment and approvals are pretty good, and OpenAI uses these tools internally. Start experimenting, keep consequential actions subject to review, and expand autonomy as you see what works. They’re patching holes, fixing bugs, and shipping features quickly.
- Growth Agent: now start prompting it and tell it what you want to do.
- Tell it to review your relevant, company-approved product, marketing, and sales data with the necessary permissions to get smart on what you or your company does, which loads up context and data into the tool. You could also have it review old resumes, your Linkedin, your X and YouTube profiles, etc to identify key areas of expertise that make you unique and well positioned in the marketplace.
- Tell it to collect the contacts you’re permitted to use from your Phone, Email, CRM, etc and build out your personal CRM or work CRM, as well as a relationship graph. This is more advanced and will take time, but suffice to say you may already have warm leads in your network you’re not paying attention to. This can help surface them. Just tell it to do that.
- Tell it you want it to go out and search the web, LinkedIn, job boards, RFP / RFQ sites, etc and look for opportunities that align with your niche expertise, or what your product or service does.
- Then tell it to automatically create a pitch, response, or message based on all your prior work in the space and the need surfaced in the market so its targeted and real.
- You can tell it you want to review it at first, but eventually you’ll probably give it an approval to send away when there’s a strong fit. Set explicit boundaries for what it can send and what needs your approval, and keep consequential actions subject to review. Again, OpenAI’s alignment standards are already pretty good.
- When you get a response back, you can set up support or response agents to tell it to check your calendars and schedule time that work for both parties including a Zoom or Teams link.
- You can tell it to create some background materials on the prospect as well so you don’t come into the call cold.
- Then you just get to show up on the call and have a warm, friendly conversation about all things human.
Let me know how it goes, and feel free reach out for any help.
—Sean