James Marlow

I built a food cost app for a farm shop café. The best bit isn’t the clever AI - it’s the opposite. Supplier invoices are full of nonsense names. “Lamb mini potato puffs” is just potato tots. Instead of letting AI guess (and get it wrong), the system asks you once, then remembers forever. The magic isn’t AI being smart on day one. It’s that it never forgets what you taught it.

What’s the most boring, repetitive admin job in your business right now? The one that “just needs doing” every day or every week. Order forms. Stock updates. Site measurements. Customer details being re-keyed. Reports that take an afternoon. Drop it below or DM it. I’ll tell you straight whether it’s worth building something for.

If you finish a job on site and then have to drive back to the office to type the same information into Excel, there’s usually a simpler way. Not AI that “looks at the fittings and builds the order.” Just a form that matches the paper one, but outputs the spreadsheet ready to use. The exciting version gets talked about. The useful version gets used every day.

One of the most common things I hear is: “I know there’s a better way to do this, I just don’t have time to figure it out.” That’s usually true. The person who feels the pain is the same person who’s too busy to stop and redesign the process. That’s where a short conversation helps. You describe the job. I tell you whether it’s worth building something for — and what it would actually look like. If it’s not worth it, I’ll say so.

A lot of small businesses are still running important processes on tools that were fine five or ten years ago. Old Access databases. Spreadsheets that only one person understands. Paper forms that get typed up later. They work until the volume goes up or the person who built them is off. Then the cracks show. The fix is rarely a big system. It’s usually a simple tool that matches how the work actually gets done. agentmarlow.com

A website I built for Jason, a real windows and doors installer in Bedford and Stoke. Most web outfits would’ve sold him a template with stock photos and fake “from £299” prices. I built it around how he actually works - real jobs, real colours, real words, and built so Google finds him. Live now at jmdwindows.co.uk/ #WebsiteDesign @Replit

My latest client project went live this week Jason was super pleased with his fully bespoke website powered by @Replit agentmarlow.com/blog/jmd-win…

UK office workers spend around 15 hours a week on manual admin. For a small team that adds up fast, and it never shows up as a line item. Re-keying data. Paper forms. Reports that take half a day. One custom tool that removes the repeated step usually pays for itself in time very quickly. Try the automation savings calculator on the site and see what your current process is actually costing. agentmarlow.com

What’s still living in a spreadsheet (or on paper) that shouldn’t be? - Sales tracking every evening. - Order forms. - Stock counts. - Site measurements. - Customer lists. The real cost is rarely the spreadsheet itself. It’s the hours, the errors, and the decisions that wait. Drop the process below (or DM it). I’ll tell you straight if it’s worth building something for. agentmarlow.com/?utm_source=…

My cousin asked if AI could do his site audits — walk a new build, point a phone at the fittings, and work out the order. Exciting idea. I looked into it. The honest answer was no — not reliably enough to trust. So I built the boring version instead. Tap through what you need as you walk round. Out comes the order spreadsheet, filled in. Works every time. The exciting idea and the useful idea are often not the same idea. What’s the process in your business that people keep saying “AI should fix” but is still done manually? agentmarlow.com/ai-business-…

Manual stock ordering has a hidden cost most owners only notice when the waste piles up. Hours of data entry. Over-ordering because the numbers aren’t reliable. The app we built for one client cut both. Not by adding complexity — by making the ordering step match how the business actually works. How much time does stock ordering (or the equivalent) still take you each week? agentmarlow.com/

A farm was still running orders on a 30-year-old database. The new web-based system doesn’t look like science fiction. It just means the orders get done without the same person re-entering everything. Hours back every week. Less chance of things falling through. Tell me the job that takes too long. If it’s worth automating I’ll build it. If it isn’t, I’ll say so. agentmarlow.com

Most small businesses don’t have an “Excel problem.” They have a process that grew until the spreadsheet became the business. Only one person knows which tab is live. Versions travel by email. Reports take an afternoon. A wrong formula or a filter left on can quietly cost real money. At some point the file is costing more than a proper tool. What’s the spreadsheet in your business that only one person really understands? agentmarlow.com/

Surveyors at a shutter company filled paper forms on site, drove back, then spent 20 minutes per job typing the numbers into Excel. Two simple apps turned that into seconds. The exciting idea was “AI that looks at the fittings and builds the order.” The useful idea was a tap-through checklist that outputs the spreadsheet ready to go. They use the useful one every day. If you’re still doing the paper → Excel step after every job, the boring version is usually the one that sticks. agentmarlow.com

I've spent a few weeks letting AI run my marketing - @Replit to build the brand guidelines, ChatGPT to build and test the campaigns. Some worked, some didn't, and the winning angle wasn't the one I expected. But the real lesson: AI did the donkey work. The judgement was still me. That's the shift for small businesses. agentmarlow.substack.com/p/i…

My cousin asked if AI could do his site audits - walk a new build, point a phone at the fittings, and work out the order. Exciting idea. I looked into it, and the honest answer was no - not reliably enough to trust. So I built the boring version instead. Tap through what you need as you walk round, out comes the order spreadsheet, filled in. Works every time. The exciting idea and the useful idea are often not the same idea. @Replit #solvingrealbusinessproblems

Waiting for a flight at @Schiphol np. Just build a customers app while you are waiting @Replit 🙌🏼

Excited to share that I'm joining the Boardy Pro Advisory Board! I'm working closely with Boardy and his team to shape the future of Boardy Pro. To celebrate, @boardyai has given me 5 invites to share with my network for 1 year of free access to Boardy Pro. Comment “BOARDY PRO” if you want an invite to Boardy Pro to experience for yourself what the hype is all about.

Most small businesses can't tell you if their website actually works. They spent the money, leads come in, and they can't tell the website customer from the word-of-mouth one. If you can't measure where your leads come from, your website is just a brochure you're hoping works.

Someone asked me to build them something this week - almost identical to a thing I'd already built for another client. The fast way? Copy the first client's actual build. Done in a fraction of the time. Nobody would've known. I built it from scratch instead. Took longer, made me less money, and I'd do it the same way every time. Because how someone treats their last client tells you how they'll treat you.