Part II

Create Your Automation Roadmap

As you move into planning your automations, remember this: your automation needs to meet your workflow where it's already at. Do not overhaul your workflow to make automation "easier." Do not replace software or ask stakeholders to change how they interact with you.

This is important, because you know your existing workflow is already valuable. Small changes can create problems in surprising ways, or introduce friction in a way that tanks your project. Sometime later you can change more variables: but initially, your job is to keep every step the same, and work surgically to delegate bits and pieces of it to your agents.

As far as any of your stakeholders are concerned, they shouldn't even know you have introduced automation.

So let's build a project plan: what are the specific steps you'll take to make progress on your automation, piece by piece?

The key here: sequence your roadmap so you are getting value at every step of the way.

Automation projects like this may take a long time. It's easy to get excited and bite off more than you can chew. Assume it will take you longer than you expect. Put yourself in a position to have small wins. Don't put yourself in a position where you're sitting on a half-done project you've spent 30 hours on, but have no productivity gains to show for it yet.

The simplest way to think about it is to categorize each of your manual work items with one of three categories:

1. Automate this now; it's not too complicated and would save me time
2. Automate this later; it's either complicated or not particularly helpful to take off my plate
3. Will never automate

The last category is tricky. Let's talk about it.

To automate or not to automate? It's about the human connection

Many specific functions we do in our jobs today will eventually be automated away by AI. But one type of function we expect to stand the test of time: those involving human connection.

We pay big money to watch NBA players entertain us, and it's important to us that they're not on performance enhancing drugs. We expect our doctors to have good bedside manner, and getting a cancer diagnosis via email will never be acceptable. Those physical world examples are fairly obvious. What's the equivalent for the digital world?

The answer: trust.

You don't believe everything you read online. Increasingly, you only trust what certain influencers have to say. Or you only trust the writing and content of teams and companies that have a storied history of reliable reporting, and background you can trust is free of biases and mal-incentives. It all boils down to trust: of an individual, or of a brand.

Examples of where trust manifests:

• Delivering a code review to your colleague that they trust is a representation of your best engineering capabilities
• Producing art that consumers trust was inspired by a lived human experience
• Writing newsletter content that your readership trusts is representative of human leaders and decision makers in your niche

Importantly, it's not "all or nothing." You can deliver great newsletter content by having AI collate candidate information, and then use your human expertise to filter and curate it. You can produce art by coming up with an idea, then using AI to put pen to paper.

But there are also some technical realities holding back AI agents today

While we believe the notion of human-human "trust" to be immortal, the on-the-ground reality of today is that there are a few more "exclusion criteria" for automation opportunities than there might be as soon as a year or two from today.

Is precision important? LLMs are inherently nondeterministic. Even if they're right 99 times out of 100, is that good enough? When sourcing news, missing 1 out of 100 stories is acceptable. But publishing even 1 incorrect blurb out of 100? Not acceptable. That's why we won't be automating final proofreading anytime soon.

Where are there technical integration gatekeepers? Not every external software will take kindly to your attempt to insert an agent in your stead. DoorDash makes a lot of money off of running ads on their UI, so they won't want you programmatically dodging them. Using automation on Craigslist is a violation of their ToS, because that's how they combat spam. In our case, our SaaS email provider does not expose a REST API and has indicated to us that automating their web UI is a violation of their ToS. Rather than fight that, we're migrating over to an alternative where no such misalignment exists.

Who has a good MCP server available? Even if there is no gatekeeper per se, the unfortunate reality is that most of the thousands of MCP servers out there today are poorly designed. If you're a product builder, you can remedy this for yourself by implementing the gaps: we're doing so for several of our SaaS tools. If you're not a builder, then the reality is that you'll want to ensure there's a popular, production-ready MCP server out there for the system(s) your workflow requires you to integrate with.

Finalizing your Roadmap

Let's revisit our target categorizations:

1. Automate this now; it's not too complicated and would save me time
2. Automate this later; it's either complicated or not particularly helpful to take off my plate
3. Will never automate; this is part of my work's core value proposition

Earlier, we presented what the automation opportunity looked like for us:

That's the end-state. More specifically, our "roadmap" initially looked like this:

Part III →