News

​Lior Weinstein: Fractional CTO/CRO, CEO, and Cofounder of CTOx, coaching tech executives to multiply their impact as fractional leaders.

gettyIn less than 68 hours this past week, the AI harness I built to run my companies merged 505 finished changes across nine products and closed 417 change requests.

This is the level of output you might expect in a push week of a 330-person engineering organization, and a U.S.-based team of that size costs roughly $250,000 a day. My model bill came to $33,500 at list rates, about $11,900 a day, and the only human involved was me.

​The implications of this for business as we head into 2027 are worth every CEO’s attention. ​

For 20 years, the business landscape was nimble startups against established, well-funded large enterprises. You either had the turning radius and implementation speed of a five-person company or the output of hundreds of engineers. Now you can have both: jet ski speed with cruise ship engines. Decide something today, and it can be live tomorrow, no matter how large your team is.

What This Means For Your BusinessI’ve previously written about how AI is compressing timelines, enabling growth without proportional headcount. If revenue doubles, your team may only grow by 30%, instead of 100% as it might have a few years ago. ​

The advancements in AI over the last four months alone have changed how I plan for staffing and growth. Earlier this year, I had forecasted needing a team of 35 to 40 people at one of my companies. Now, however, I have extreme confidence I’ll actually need less than 20 people.​

When the capacity to build is there, you also compress time to market. A change from my harness is finished and goes live the same day. A larger organization might take a few days for the same change based on their human handoffs and review queues.

Why Progress Still Feels SlowSo if the constraint is no longer development speed, why does AI adoption still feel so slow?

Because as a business community, we haven’t solved the primary constraint: consensus.​

Examine the path an initiative takes in your company to become real. First, there’s an idea. Then, some discussion. Someone says yes or no. Someone refines it, and it eventually goes into a queue to be developed. That path is a loop, a human-powered algorithm. ​

For decades, we didn’t measure the efficiency or effectiveness of this loop, because it took so long for ideas to be built and developed. Now, it’s the only math that matters for a human team.​

When the development team is a single operator managing an AI harness, like mine is, the trade-offs of human teamwork are real. Adding even one additional world-class, highly capable human would slow down my system. One more person expands the loop and requires explaining and polishing before building. The fewer people you have, the faster it is to get consensus.​

Fred Brooks wrote about this over 50 years ago in The Mythical Man-Month: Adding humans to a late software project makes it even later, because the communication paths multiply faster than people. Bain & Company found that each person added to a decision group beyond seven in size reduces decision effectiveness by about 10%.​

In the 68-hour window I mentioned above, 39% of the time was spent deciding what to build. The actual step of building took 20% of the time. With AI, as with human teamwork, deciding costs more than doing.​

Just as the idea-to-implementation process is an algorithm, so too is consensus. The number of people who have to sign off creates its own structure, which creates predictability and stability but also introduces bureaucracy. As a business owner, your job is to evaluate all the bottlenecks across the organization and sift the ones you’re willing to tolerate and the ones you’re not.

Where Human Attention GoesIf you have fewer people in these ideation loops, what do they focus on? In many organizations, fewer people leads to layoffs. ​

My recommendation is elevation over elimination. Humans who leave the idea-to-implementation loop can refocus their time and attention on front-stage work, focusing on the experiments and novel ideas only the person in that seat has.​

This is where leadership adds unique value through vision. You’re ideally positioned to help your team see the gifts they have that AI can’t automate and involve them in the evolution of their role or the department. Some will need coaching along the way. They may be carrying wrong assumptions about where they create value.​​

Four Ideas To Put This Into Action1. Count the nodes and loops: Look at the core processes inside your company, and count the humans between an idea and the build queue for each.

2. Sift and sort: For each person in the loop, determine whether they are there for judgment or comfort. The test is simple: Who makes the daily decisions, and who is in the room so nobody feels left out?

3. Limit each AI build to one decider: One accountable name who says yes or no and refines before anything enters the queue.

4. Communicate the change: Tell your team where the seats leaving the decision-making loop are going and what those people will focus on instead. ​

How many people stand between your next idea and your build queue, and which of them are there for comfort?​

Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?