Writeup
meta-muse-wonderrun 13 Oct 2026196,341 tokens
Instructions
Instructions — what the model was asked
Hey, I'm trying to write a post on my experience with Every Consulting. The consulting so far, I just joined as the head of tech consulting. I didn't tell you who runs consulting. I'm going to provide a transcript of me talking about my experience. I was also going to provide a tweet, which is just a list of the learnings that I've had so far for companies that were struggling to adopt AI. It...
Hey, I'm trying to write a post on my experience with Every Consulting. The consulting so far, I just joined as the head of tech consulting. I didn't tell you who runs consulting. I'm going to provide a transcript of me talking about my experience. I was also going to provide a tweet, which is just a list of the learnings that I've had so far for companies that were struggling to adopt AI. It was in response to Sam Par, who was asking, "How is everyone getting team adoption for Claude? How do you get your team to actually use it effectively without spending all their time on Twitter and learning?"
I need to write this in my writing style. I've included some writing of my own as a sample. I also have included a post on consulting at Every, so from in Natalia, you get a sense of what consulting looks like at Every. But yeah, focus mostly on the specifics of the interview that I provided with the context of that tweet.
https://x.com/hammer_mt/status/2032591631413567873?s=20
Learnings:
- buy the model direct not 3rd party tools
- raise the ceiling not the floor
- workshops with 50%+ build time
- assign impossible tasks
- mandatory AI note taking + MCP connector
- map workflows and systematically automate
- train people to be managers (of agents)
In response to:
https://x.com/thesamparr/status/2032477081695584670
How is everyone getting team adoption for Claude?
I spent a lot of time on Twitter, as do you. We see all this AI stuff popping up. We're on top of it, or at least sorta. I know what's going on and are testing all these fringe ideas.
But how are all you people getting your team to actually use it effectively without spending all their time on Twitter and learning, which we know they won't and probably shouldn't be?
What Actually Gets Teams to Use AI
Seven lessons from Every Consulting
Author's note: Written as a draft post in first person as head of tech consulting at Every, based on the transcript and learnings list provided. Two assumptions are made where the source material was incomplete: (1) "who runs consulting" was deliberately not stated in the prompt, so the post refers to Every Consulting generically rather than naming a principal; (2) "Natalia's post on consulting at Every" was mentioned but not included in the files, so the framing of consulting at Every is drawn only from the transcript's consulting-team and discovery-call anecdotes, not invented specifics about Every's consulting practice.
Sam Parr asked a question [on X] that I keep hearing from executives: How is everyone getting team adoption for Claude? How do you get your team to actually use it effectively without spending all their time on Twitter and learning, which we know they won't and probably shouldn't be?
I spend a lot of time on Twitter, so I know what's going on and I am testing all these fringe ideas, at least sorta. Most teams are not going to do that. They shouldn't have to.
I joined Every as head of tech consulting, and the consulting work so far has given me a clearer answer than anything I had from experimenting alone. These are my learnings for companies that are struggling to adopt AI. None of them is "buy better prompts." They are mechanisms for giving direction to AI and human coworkers at the same time.
Every manager is a prompt engineer, using their communication and data analysis skills to align the team toward a common set of goals. Even your employment contract is a prompt: a standardized template of language designed to align your behavior with the commercial goals of the organization. AI adoption works the same way. Those who get good at this set of skills will have an unfair advantage over those who don't. Let’s dive into the seven principles that actually stuck.
1. Buy the model direct, not third-party tools
People will go and evaluate these tools that are using Claude or Codex or Gemini under the hood. The provider has its own hang-ups and product decisions and opinions baked into the tool, and quite often it's quicker and easier just to create a Claude skill that has your own opinions and your own subjective approach to work directly.
There is a real move towards build, not buy, just because it gives you flexibility. It's much easier to configure and operate your own.
I don't know how it's possible for companies that are not the core model companies to keep up. The core model companies know what releases are coming. They build their internal tools to schedule those releases. They train, for example, on how to operate within the Claude Code environment. I appreciate the effort that companies like Cursor put in to keep up—they're doing a great job, they're a good product org—but I don't know how you compete with Anthropic offering $5,000 worth of tokens a month for a $200 a month subscription.
Third-party tools tend to be less flexible, less on the cutting edge, and more expensive. That's not always the case, but as a general rule, it makes sense.
2. Raise the ceiling, not the floor
A lot of people come in and say, you all need to use AI now, we bought you AI tools, you don't need to adopt it. I don't think that works.
Even on pain of death, a lot of people are just unwilling emotionally to use AI or be told that they have to use AI. It's a self-preservation thing. Instead you need the carrot rather than the stick.
Nominate people who are already AI-forward. Make sure it's understandable that it's encouraged to use AI, and get people out of the woodwork—people who are using Claude Code in their spare time—and just give them the support they need. Someone who's really AI-pilled is going to be able to do something like five, ten times more than someone who just does not agree with AI yet, or hasn't seen the magic yet.
People need to be able to have that moment on their own time, but you can accelerate it by showing that people who use AI aggressively are the ones that get promoted first, the ones that get time with senior management. We've even seen some success with co-opting those people to be TAs in the courses that we're teaching to the rest of the team. When they see that person is accomplishing more and getting ahead in their career and getting more face time with management, that's a more effective motivator.
You also get the productivity boost of just enabling that person who's already a believer. It's much harder to convince someone to believe.
3. Workshops with 50%+ build time
You should do workshops, first of all. But workshops should be at least 50% build time.
People do not want to sit there and look at slides and theory. The way that I learned how to do AI was by just doing. You need some guided theory to motivate people to do stuff, but the biggest problem we heard from people is they just didn't have time to check out these tools. They didn't have time in their day to learn something new and push themselves and stretch themselves.
If you give them a couple of hours in a workshop where they're expected to build something new and go outside their normal domain, and they have access to the tool and the facilitator has given them some synthetic data to work with or made sure they had access to the connectors they need, that's where they get that aha moment and they really get it and they start working on it.
Direction is rarely enough on its own. You have to specify format, provide examples, and divide labor—but none of that lands until someone has built something themselves.
4. Assign impossible tasks
What I mean by impossible tasks are things where it just wouldn't have been possible to do that task without AI.
Boris Cherny, the inventor of Claude Code, has said something along similar lines: you should slightly under-resource most teams, and that makes them think, okay, well, the only way I can do this is if I use AI.
I like to make it more explicit and strategically choose things. If your goal is to write one blog post a week, you can just do that manually. If your goal is to write one a day, you're going to have to use a lot of AI in the research phase, in the editing phase, etc.
You don't set the goal as, starting today you have to produce one a day. Instead you set the goal as, our goal is to get to the point where you are producing one a day—what needs to happen, what's your progress towards that goal? It might take some time to get there, but if they know that's the goal, they'll start thinking strategically about how we can use AI to save time and start experimenting.
5. Mandatory AI note taking + MCP connector
This is a really big unlock.
Everyone in the consulting team has Granola and the Granola MCP. It's just so much easier when you have that data at your fingertips to be like, okay, we just had a meeting on X, create an update email on that topic and then send it to Y.
That is 80–90% of the value of AI, to be honest: being able to extract information from unstructured data and structure it in a way that's useful. There are so many times where I've come to a task and realised I need some context from a meeting that we had, and now we can just pull it from the MCP. I always include that information when I'm creating curriculum, when I'm putting together proposals, etc. It's really useful.
Just like you would brief a writer on word count or structure, you need to give the model the context it can't count on its own—and meetings are where most of that context lives.
6. Map workflows and systematically automate
We have a process for doing discovery calls where we talk about what they do day-to-day, what tools they use, what things they do on a daily basis, what are the pain points. Then we turn that into a structured Google Sheet where we have a list of the different things we need to solve for, and then we just systematically work down that list.
The assumption is that we're trying to get 100% of the stuff off the people. At the very least, AI is doing the first pass of these, and we have a skill for each of the type of tasks that you're working on. Then that person could be doing five to ten times the throughput that we're doing right now.
So far, that doesn't seem to ever lead to people being let go. In fact, what it leads to is either they put way more effort into each task, or they just manage to expand the revenue of their team or the throughput of their team without having to hire new people.
An example of investing way more in these tasks: when we were doing courses for the team, we would prepare one task for the overall cohort. Maybe we could prepare one task per team pod. Now we're actually at a point where we could use Claude Code to create an individual project for each person that's taking part. That wasn't possible before AI. So we're using AI to put way more effort into each individual course rather than firing someone or automating work away.
7. Train people to be managers (of agents)
Everyone who was an individual contributor before is now a manager, and they're struggling with the fact that they don't have management training. They're managing these AI tools, they don't know how to do it, they're not used to context switching, they're not used to setting up systems or rules in place, they're not used to evaluating or having to give strong opinion about whether something is good or not that they haven't actually worked on themselves.
Managers can take to this more readily once they have that aha moment than individual contributors do. Individual contributors enjoy the craft and care about how something is done, whereas managers have long since gotten past the point where they care that much about how a problem is solved—they just want the problem solved to the specification that they require.
At the same time, the funny thing is that managers now are becoming individual contributors because they can manage a team of agents more effectively often than they can manage their human teams. It's lower bandwidth, sometimes, to brief a human, and it's a longer feedback loop. So it's actually easier sometimes as a manager to vibe code a task, use Claude Cowork to do a task, than it is to brief a human to do it and the human uses Claude Cowork. By the time you've written the prompt, you can decide whether to send it to Claude and get the response in five minutes, or send it to a human and then wait for them to send it to their Claude and get a response in a couple of days.
You need more management training for these companies. You need to help them understand context switching. You need to teach them how to do evals. You need to teach them how to develop good taste—deciding what to work on. You need to teach them how to train the AI in specific skills: how do you systematically write a really good PowerPoint skill? How do you systematically write a really good daily update report skill for Claude? And work through those things.
That's how I think about it.
Links referenced: Learnings post in response to Sam Parr, https://x.com/hammer_mt/status/2032591631413567873 — Original question, https://x.com/thesamparr/status/2032477081695584670