AI training for a leadership team: the challenge
Connectd had adopted basic AI fast and then stopped progressing. The whole company was mostly using ChatGPT as a chat box.
When the leadership team signed up, all eight reported the same thing: no agents running in the background, and most had never experienced an agent working on a task for longer than five minutes.
The solution
Roei bought seats for himself and his senior team, and mandated the training time: three hours on a Tuesday, three on a Friday, an hour on a Thursday, for four weeks.
He realised that uneven skills across a leadership team produce uneven outcomes, so the CEO, COO, CCO, CMO, VP of Sales and three function heads all went at once.
By the end, six of the eight had shipped something running against live company data, including two custom MCP servers built by people who are not engineers.
The Connectd team created agents that work across
How was Connectd’s team using AI before the Agent Accelerator?
We were very quick at adopting level-one large language models, using ChatGPT for everything. We have a young team, so nobody was opposed to it. But because of the speed Connectd is growing, and how much gets done on the fly, it is hard to take a step back. We were strong at adopting that first iteration and getting it useful. I do not think we ever really evolved from it in a meaningful way.
When Cowork came out I started playing around with it, but I did not really know how to use it effectively. I connected a couple of things and thought, I can send a Slack message through Claude rather than through Slack. So I had added a step to sending a Slack message. There was obviously a better way, and we could not figure out what it was.
As a founder and CEO I do not have much manual repetitive work, so there was no obvious low-hanging fruit for me. I knew there would be amazing use cases for the team, but I could not work out how to make that happen.
“I knew that there would be amazing use cases for the team, but just couldn’t really figure out how to make that happen.”
How did the team feel about AI before the training started?
The willingness was there. The time and the expertise were not. And you get a lot of AI paralysis and anxiety. Every single day you are on Instagram and somebody else has the perfect stack and there are fifty new ways of using it.
Everyone I speak to feels far behind. I feel behind. And that feeling makes it harder to engage, especially if you are still at step zero or step one.
What convinced you to buy eight seats an hour after you discovered the Accelerator?
The post linked to a testimonial from a founder, a CEO, talking about how much it had unlocked AI for their team. That made sense to me.
The other thing is that the only way to make training happen is to force it. Everyone is busy, rightly so, it is a high-performing team. Unless you say we are going to make time, it does not happen. So it was: you have zero notice, this starts in a week, it is three hours on a Tuesday, three hours on a Friday, an hour on a Thursday and plenty to do in between, and it just has to happen. Everyone was on board, and it was brilliant.
“The only way to force functional training is to force it.”
How would you summarise the Agent Accelerator to someone considering it?
The best way to summarise the value is that I had one team member who went from using a large language model as a search function, or to summarise text and emails, to building custom MCP servers within four weeks. Which is insane.
A lot of it is credibility. There are so many charlatans in the world of AI, so people really listen to you, and that gives them confidence that this is the right way to use these tools. There was a strong emphasis on doing it safely too. I am a risk-averse founder, and that gave the team a lot of confidence.
“I had one team member who went from using an LLM as a search function or summarising text or emails, to building custom MCPs within four weeks. Which is insane.”
As a founder and CEO, how has your approach to AI changed?
I now use Claude primarily as an executive assistant and almost a chief of staff. As a founder with ADHD, I did not realise how much I was missing. You come out of a two-hour conversation with an exec with a fairly clear picture of the actions, but there was a lot I was missing.
With the file structure we built, which is one of the biggest changes from a usage perspective, I have that level of context in one place. I have proactive agents that take my daily notes, everything from Slack and everything from email. Being able to now go back and forth with true context is a game changer.
Every day I get digests that take account of unresolved items and action items from calls I might have missed before. This weekend I am planning to build a best-record skill, so if we onboard a record number of startups in a month it drafts the investor update from our own records. That is not manual repetitive work, and I am still getting value from it.
“Being able to now go back and forth with true context is a game changer.”
What impact did AI agent training have on the team?
People feel really empowered now. I think that is the blanket feeling, because people were not using it anywhere near its capabilities before.
It has changed the way we think about headcount expansion, particularly in some functions, in customer and in parts of marketing. We are about 150 people now, which is a good size, but I believe we will be able to empower those people to have the impact of 250 or 300, especially as the tools get better adopted and we work through the data project.
“It’s changed the way we will think about headcount expansion.”
How many people should a company send on AI training?
So much of it is context. I knew that if I had one of my leaders on the course and they started wanting to use AI in a different way, and you have uneven skills, desire and understanding across the leadership team, it just creates uneven outcomes. Which is not what you want.
So if you are serious about doing this, take the plunge. Send all of the right people, all the right leaders. For us it has been unbelievable.
“If you’re serious about doing this, take the plunge. Send all of the right people, all the right leaders. For us, it’s just been unbelievable.”
What would you say to a leader on the fence about the Agent Accelerator?
Every business owner has the fear that they are falling behind. Every business owner does have that fear that they’re falling behind. It’s like not going to the doctor when you think something’s wrong. Getting the information, getting the results and seeing where you actually are is the hard part. Once you are there, it is easy to make meaningful change. So bite the bullet, do it, and you will feel empowered on the other side.
“Every business owner does have that fear that they’re falling behind. It’s like not going to the doctor when you think something’s wrong.”
What they built
A lead digest for six account managers, and an NRR skill on a custom MCP
Jamie Head of Customer and Startup
Six account managers no longer start the day working out which leads need chasing. Jamie wired HubSpot to Slack so the list arrives every morning, split into new and existing, with last-activity dates on the admin items. He then took a metric he is measured on, fulfilled net revenue retention, which had meant pulling from two separate databases by hand, and turned it into one skill by writing a custom MCP server to join them.
“This should save the team time every day and make it really easy for them to keep on top of leads they have been assigned. Now that the skill knows how to judge fulfilled NRR I can also iterate on this for other cases such as renewals forecasting and planning in Google Sheets.”
A morning CEO digest, running on a schedule
Roei CEO
Roei was losing action items out of long conversations with his execs. Now three scheduled agents catch them: one reads the tactical and exec Slack channels at 8am and writes up customer risks, wins, blockers and a hiring pulse; a second chases four named execs every Monday for their update; a third assembles a week-at-a-glance brief on Sunday evening from the calendar, the finance channel and Drive.
“Every day I am getting digests, which take into account unresolved items, action items from calls that maybe I was missing before.”
A decisions log that writes itself every morning
Lynn Head of Brand and Content
A weekly job Lynn used to do by hand, and sometimes dropped, now takes thirty seconds and runs every morning. It reads three Slack channels and all her direct messages, works out where someone is waiting on a reply from her, and writes a decisions log into Notion organised by priority with the open items flagged.
“Previously this was a manual task and I dropped the ball on a few things. Now it takes 30 seconds and I can run it every morning.”
Roadmap updates, from Notion to a finished deck
Claudia COO
Claudia was spending hours a week telling stakeholders where the roadmap had got to. The skill she built live-queries the revops and customer-ops roadmaps in Notion, summarises them to six slides, builds the deck and then visually checks its own output, so each stakeholder still gets their weekly update and she assembles none of it.
“I sink in hours a week updating people on our roadmap items, now I do not have to.”
Put your own leadership team through it
Cohort 4 starts Tuesday 29 September 2026. Four weeks, live, with the whole team in the room at the same time. You can read the four-week curriculum Connectd’s team went through, or 119 reviews from the people who have done it.