AI implementation at the desk level. What the build actually looks like, not what the strategy deck says it should. This is for the person who has read everything about AI and still doesn't know what to do on Monday morning.
Sign up and get 3 free frameworks ↓The AI Audit Framework
The exact 5-step process I use to audit a company for AI readiness. Not the version in the pitch deck.
The Question Nobody Asks
Everyone asks which AI tool. Nobody asks which task to give it first. The tool is not the decision.
What Urban Crave Taught Me About Timing
I lost my hospitality business to timing. Waiting one quarter too long. That experience changed how I work with every client.
The AI Agent Trap
A founder spent $80K on a custom AI agent. It works perfectly. Nobody uses it. This is not a technology problem.
I audit companies for AI readiness. Here's the exact 5-step process I use, not the one in the pitch deck.
I audit SaaS companies for AI readiness. Here's the exact process I use:
Most companies want to start with Step 3. The ones who skip Steps 1 and 2 are the ones who come back in six months having spent the budget and changed nothing.
Most consultants build the system first. Then spend months convincing people to use it. I do it differently.
I've never had a team resist an AI tool I deployed. Not once.
Most consultants build the system first. Then spend months convincing people to use it. I do it differently.
So by the time we're done, the team isn't adopting a system I built. They're defending a system they helped build.
That's the difference. Ownership kills resistance. Every time.
Uber exhausted its entire 2026 AI budget by April. High usage is not the same as high ROI. Three questions only.
Uber exhausted its entire 2026 AI budget. By April. 5,000 engineers. Usage doubled in 8 weeks. 70% of code now written by AI.
When I read this, I wasn't surprised. I see a version of this every week. Not at Uber's scale. But the pattern is the same. A company buys a tool. The team adopts it. Nobody tracks what it's actually producing.
High usage is not the same as high ROI.
Before I recommend any AI tool to a client, I insist on one thing first. Three questions only:
If you can't answer all three before you buy, you're not buying a solution. You're buying a problem you'll need to explain to your CFO in six months.
The companies spending most on AI are the slowest to change. The ones who moved quietly started somewhere no consultant ever suggested.
The companies spending the most on AI are the slowest to change. The ones who moved quietly started somewhere no consultant ever suggested.
Not the boardroom. Not the strategy offsite. Their own desk.
That is the whole sequence. One person changes one task. The team notices. The organisation follows.
It does not start with strategy. It starts with a 30-minute decision at someone's desk.
The sequence is Desk → Team → Organisation. Never reverse it. Every AI rollout that started at the org level and tried to reach the desk eventually failed. Every rollout that started at the desk and spread upward succeeded.
The reason is simple: desks produce evidence. Boardrooms produce plans.
A client uses AI every day for her own work. Her team thinks she is sceptical of it. She hasn't told them yet.
A client uses AI every day for her own work. Her team thinks she is sceptical of it. She has not told them yet.
That gap between what leaders do privately and what they say publicly is where most AI adoption is actually happening right now. Nobody is waiting for permission. They are just not talking about it.
The most honest signal of where AI is in your organisation is not the policy document. It's what your senior team quietly uses before 9am.
I sat in a meeting this week that AI could have replaced entirely. The meeting wasn't the problem.
I sat in a meeting this week that AI could have replaced entirely. Not because the conversation was useless. Because the preparation wasn't done. The brief wasn't written. The data wasn't pulled. Three people's time spent catching up on what one person and thirty minutes of AI work could have done the night before.
The meeting isn't the problem. What happens before it is.
Most "too many meetings" problems are really "not enough preparation" problems. AI fixes the second one. And when the second one is fixed, the first one mostly solves itself.
Everyone asks: which AI tool should I use? Nobody asks: which task should I give it first.
Everyone asks: which AI tool should I use? Nobody asks: which task should I give it first?
The tool is not the decision. The task is. Pick one thing you do every day that is mostly about processing or producing information. Start there. The right tool becomes obvious after that.
The question never needed to come first.
Three months. That is how long the AI pilot sat in procurement. The team had already built a workaround.
Three months. That is how long the AI pilot sat in procurement. The person who requested it had already built a workaround using the free version. The team had been using it quietly for six weeks.
By the time the budget was approved, the adoption had already happened without it. The approval process didn't enable the transformation. It just documented it.
This is the most common pattern I see in organisations above a certain size. The people who need the tool don't wait. The process that's supposed to give them the tool lags six to twelve weeks behind reality.
Most companies measure AI adoption by licences purchased. Nobody measures it by tasks actually changed.
Most companies measure AI adoption by licences purchased. Nobody measures it by tasks actually changed.
One person on your team who has genuinely replaced a daily task with AI has delivered more value than fifty people with access who haven't touched it.
The number that matters is not how many seats you bought. It is how many habits changed. Until you're measuring that, you're not measuring AI adoption. You're measuring a subscription.
Every AI strategy I have seen fail had one thing in common. It started with a deck. Not a task. Not a person. A deck.
Every AI strategy I have seen fail had one thing in common. It started with a consultant deck. Not a task. Not a person. Not a desk. A deck.
Forty slides about transformation that ended with a roadmap nobody owned.
The companies that got it right started differently. Someone changed how they did one thing. Then told the person next to them.
The deck is not useless. But it should be written after the desk has already started changing. Evidence produces strategy. Strategy alone produces more strategy.
AI wrote a brief for me last week that was technically correct and completely wrong.
AI wrote a brief for me last week that was technically correct and completely wrong.
It had the structure. It missed the nuance. It didn't know what the client actually needed to hear versus what they asked for. That gap between what was asked and what was needed is years of experience.
AI showed me exactly where my value lives. I wasn't expecting to find it that way.
What AI gets right is not your competitive advantage. What it consistently gets wrong, and you correct without thinking, that's the list worth knowing.
I spent 20 years building businesses without AI. Scaled a family enterprise to $100Mn. Built a central kitchen brand. Then I saw what AI actually does inside a business, not a demo, not a pitch deck.
I spent 20 years building businesses without AI. Scaled a family enterprise to $100Mn revenue. Built a central kitchen brand from scratch. Launched ventures across 7 industries. Operated with 700+ distributors, 1,600 MT/day production. All of it, manual. People-dependent. No automation.
Then I saw what AI actually does inside a business. Not a demo. Not a pitch deck. I sat with a CTO. Audited every process. Mapped every workflow. And rebuilt the product as an AI factory. That changed everything I thought I knew about running a company.
Here's what I learned: most companies don't have an AI problem. They have an execution problem. They buy tools. Hire consultants. Run pilots. And nothing changes at the desk.
When I launched my own practice, working across organisations in the United States and India, the same companies came back. Not as colleagues. As clients. Because the build is where it lives. Not the strategy.
My client was tracking 47 videos across 12 data sources. Three people. Spreadsheets. Copy-paste. They had tried adding AI twice. Both times it stalled.
87% of AI projects never make it to production. Last month I saw why.
My client was tracking 47 videos across 12 data sources. Three people. Spreadsheets. Copy-paste. Every week, someone pulled the wrong number. Every week, someone missed a deadline. They had been saying "we need a better system" for a year. They had tried adding AI twice. Both times it stalled.
I did not start with AI. I started with one question: what is the workflow that breaks everything when it fails? Content production tracking. Data everywhere. No single source of truth.
We built a single dashboard. Every data source feeding one place. Automated. The team spent two hours testing it. Then they stopped using the spreadsheets. Not because we told them to. Because the dashboard was easier.
The AI wasn't the solution. The workflow clarity was. AI was just the tool that made the clarity stick.
I lost my hospitality business to timing. Waiting one quarter too long. That experience broke something in me. And fixed something bigger.
While you're still evaluating AI tools, your competitor just deployed one. I know what it costs to be on the wrong side of that sentence.
2020. I watched my hospitality business die in slow motion. Competitors who had invested in delivery tech survived. I hadn't. Not fast enough.
I didn't lose Urban Crave to bad food or bad service. I lost it to timing. Waiting one quarter too long. I kept every employee on payroll through the pandemic. Paid salaries when there were no customers. Then shut it down. 8.5 years. Gone.
That experience broke something in me. And fixed something bigger. Now I help companies ship AI before they need it. Not because the technology is exciting. Because the cost of waiting is something I've already paid personally.
In 2020, Japan was filing COVID-19 case reports by fax. Not because the workers couldn't use computers. Because the structure required a physical stamp.
In 2020, Japan was filing COVID-19 case reports by fax. Not because doctors couldn't use computers. Because some systems required a physical stamp as proof of authority. The workers were ready. The structure wasn't.
When Japan's Digital Agency tried to phase out fax machines, it received 400 formal objections, from government ministries. Not from frontline workers. From leadership.
This is not a Japan story. Last year, a spice manufacturer in Jaipur told me: "When my children join the company, we can try." That same conversation, he mentioned he uses ChatGPT to draft emails to overseas buyers. He wasn't resistant to AI. He was resistant to the idea that the structure could change.
The team is almost never the problem. The structure that asks them to stamp everything before it can move. That's the problem. Fix the structure. The team follows within weeks, not months.
Not by building something new. By removing what they didn't need. The fastest way to grow isn't always to add.
A US-based SaaS company ($8M ARR) asked me to help them "add AI to their product." Before I added anything, I audited what they already had.
What I found: 4 SaaS tools doing overlapping jobs ($87K/year wasted). A manual QA process eating 60% of 3 people's time. Customer support responses being written from scratch for questions asked every single day.
When I ran a $100Mn cattle feed business, I learned something counterintuitive: the fastest way to grow isn't to add. It's to subtract. We removed 3 layers of middlemen from our distribution network. Revenue didn't drop. Margins doubled.
Same principle with AI. We didn't add new AI. We consolidated the tools, automated the QA, and built a response library. $340K/year back in 90 days, before we built anything new.
That is what I paid for AI this month. For the first time, expertise is not behind a paywall, a geography, or a connection.
Less than my mobile data plan. That is what I paid for AI this month. Less than what I pay to stay connected.
And for that, I get to work alongside something that has read everything, solved everything, and forgotten nothing. I have been working in this space long enough to stop being surprised by the technology. But the price still catches me.
Because it means, for the first time, expertise is not behind a paywall, a geography, or a connection. A first-generation entrepreneur in a tier-3 city and a well-funded founder in a metro now sit at the same intelligence layer.
The difference between them is no longer resources. It is the habit of using it. The individual who builds that habit changes their growth trajectory. That is the real story of AI in India, and nobody is writing it yet.
A founder spent $80K on a custom AI agent. It works beautifully. Nobody uses it. This is not a technology problem.
I talked to a founder last week. He'd spent $80K on a custom AI agent. It works beautifully. It does exactly what the spec said. Nobody uses it.
Not because the team is resistant. Because the agent solved a problem the team had already worked around. By the time it launched, the workaround had become the process.
This is not a technology problem. This is a sequencing problem. You cannot build the solution before you've sat with the people who have the problem. You cannot spec the agent before you've watched the workflow break in real time.
Everyone's building AI agents. Most will be ripped out within 18 months. Not because the tech failed. Because nobody mapped the desk before they mapped the deployment.
The person taking your job isn't smarter. They're running AI while you aren't.
Every conversation about AI and jobs lands in the same place: will it replace us? Wrong question. The replacement isn't coming from a machine. It's coming from the person in the next seat who figured out how to use the machine while you were debating whether to try it.
I've watched this play out at the desk level across industries. The people who fall behind aren't the ones who refused AI. They're the ones who waited for someone above them to decide it was safe to start.
The gap isn't technical. It's behavioural. One person changes one task on a Tuesday afternoon. Six months later, they're doing the work of three. That's not a statistic. That's a person I've sat across from.
Every organisation getting results started with one desk, not a strategy offsite.
I have sat through a lot of AI strategy presentations. They are usually well-designed. They always begin at the org level. They almost never mention the desk.
The companies I've seen actually change how they work didn't start with a roadmap. They started with one person who changed one task. Then another person saw what was possible and tried something adjacent. The rollout came later, after the behaviour had already shifted.
In India, I've watched solopreneurs outpace teams of ten because they had no choice but to start at the desk. No budget for a pilot programme. No committee to approve a tool. Just a task, a prompt, and the willingness to run it again until it worked.
The boardroom can fund the transformation. It cannot start it. The desk always goes first.
Free Resources
No form. No funnel. Just download and use. These are the frameworks I share with clients before we start work together. After these three, you will know exactly where your build starts. Before you spend a pound on anything.
The three questions I ask every founder before they spend a single pound on an AI agent build. Skipping these is how $80K disappears.
The exact framework I use to audit a company for AI readiness. Map it against your own operation and you'll know where to start before you buy anything.
A working template for auditing an AI deployment before it goes live. Use it to catch the gaps that cause 87% of projects to stall at the production stage.
A note, if you've made it this far
If you recognise your organisation in any of what you just read — the approval loop, the standup that says nothing, the tool nobody uses — that recognition is not a coincidence. It is the diagnostic.
Most people who come here have already tried something and watched it stall. Not because they chose the wrong tool or the wrong team. Because nobody mapped the desk before they mapped the deployment.
I work across organisations in the United States and India. The pattern is the same in both. The companies that move are the ones that start somewhere specific — one person, one task, this week. The ones that wait for the right conditions keep waiting.
Stop reading about it. Start building it. I'm here.
Everything on this page is designed to show you what's wrong and how to fix it. But at some point the reading has to stop and the build has to start. Every week it doesn't, the workarounds compound. It starts with two hours at one desk.