Hayat Amin on putting AI to work without losing human judgment
Hayat Amin, founder and CEO of Beyond Elevation, places experienced operators in growing technology businesses as fractional CFOs, Chief IP Officers or AI operations leads, with AI handling part of the work. Amin explains how the firm combines finance, intellectual property and AI, and shares what has and has not worked in day-to-day AI use.
Start with one process that costs real hours, give an AI agent a named person to check its work, and learn from that before expanding. Use AI for research and routine tasks, but keep the human involved when a message depends on trust.
What’s the quick origin story of your business, and what makes what you do genuinely different from the other options your customers are looking at?
I spent twenty years in the C-suite of high growth technology companies, mostly as CFO and Chief Strategy Officer. I ran the sell side three times, including Cake Technologies to American Express and Tripbod to TripAdvisor, and I was part of three FT100 businesses along the way.
What I kept seeing was the same gap. A company growing fast needs someone who has sat across the table from a buyer, but it can’t justify that person five days a week. So it hires too junior, or it hires an adviser who writes a report and leaves.
Beyond Elevation fixes that. Every person we place has done the job for real, inside a company that exited. They join the team, not the board pack. The part I think nobody else has right is the combination. Most firms sell finance, or IP, or AI. I think they’re one problem now. The data a company sits on, the patents it never priced, and the agents that run its back office all decide what it’s worth on the day someone makes an offer. We work on all three with one accountable person.
Where does most of your new business come from today, and where do you wish it came from?
Today, mostly from people. Referrals from founders we’ve worked with, and conversations that start on LinkedIn and turn into a call. I’m grateful for that, because a referral arrives with trust already built.
The weakness is that it depends on who happens to be talking about you that month. I’d like more of it to come from people who have already read how I think. A CEO who has read my view on why most companies are giving their data away inside a revenue multiple, or why finance teams should be built for the buyer you don’t have yet, turns up to a first call with the real question ready. Those are the best conversations I have.
So I write a lot more than I used to, and I write it to answer the questions founders actually type, not to sound clever.
What’s one thing you’ve actually handed over to AI in the last year, and one thing you tried it on and went back to doing the human way?
Handed over: most of my executive assistant work. I built an AI assistant, Clara, who runs scheduling, inbox triage, travel enquiries and reminders across WhatsApp and email. She drafts, chases and books. She also has hard rules she can’t break, like taking instructions from me only, because an assistant that anyone can talk into acting is a security hole, not a convenience.
Taken back: anything that asks another person for something. I let AI write a batch of introduction emails to people I wanted to work with. They were polished, well structured and completely wrong. One of them got a one word reply: "No." It read like a sales sequence, and the people I was writing to are peers, not leads.
Now the machine does the research, and I write the ask. A person can tell when nobody was in the room when the message was written, and they’re right to mind.
Have you noticed a change in how customers find you or what they already know before they reach out? Is anyone arriving through AI assistants like ChatGPT yet?
Yes, the change is real, though I want to be careful not to oversell it. I track where every booking comes from, and AI assistants aren’t yet a channel I can put a number on with confidence.
What has changed is how much research happens before anyone gets in touch. Founders now ask ChatGPT or Perplexity "who could help me prepare for an exit" or "what does a fractional CFO actually do" before they ask a friend. If your name isn’t in the answer, you’re not in the shortlist, and you’ll never know the shortlist existed.
So I’ve treated it like any other part of the business. My background, my exits and what I work on are written plainly in one place, in the same words everywhere, so an assistant can repeat it accurately. And I publish answers to the questions buyers actually ask. It’s the same discipline as getting a company ready for due diligence. The numbers have to say the same thing wherever someone checks.
If you were advising someone in your industry on all of this for the next 12 months, what would you tell them to actually do, and what would you tell them to ignore?
Do: pick one process that costs you real hours every week, the month end close, investor reporting, lead research, and put an agent on it with a named person who checks its work. AI that nobody watches drifts. It invents a number, nobody notices, and the board pack is wrong. One process done properly teaches you more than ten pilots.
Do: price what you own. Most companies I look at have data and know-how that a buyer will value, and they give it away for nothing inside a revenue multiple because nobody wrote down what it is. That’s a finance job as much as a legal one.
Do: make yourself easy to find for the machines your buyers already use.
Ignore: the headcount maths. Anyone telling you to cut half your team this year because of AI hasn’t run an agent against real numbers in a real company. Ignore the tool of the week, too. The advantage isn’t which model you use. It’s whether someone senior owns what it does.
Thank you to Hayat Amin and the team at Beyond Elevation for sharing what actually worked with Leaders Perception readers.
Want to share your playbook?
Leaders Perception publishes founder and operator interviews every week. A few questions, about five minutes, no calls. It is free and everyone we accept gets published.