Alex

[Alex @AgenticSoul]

𝗜𝗳 𝘆𝗼𝘂𝗿 𝗔𝗜 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻 𝗶𝘀𝗻'𝘁 𝘀𝘁𝗶𝗰𝗸𝗶𝗻𝗴, 𝘁𝗵𝗲 𝗳𝗶𝗿𝘀𝘁 𝘁𝗵𝗶𝗻𝗴 𝗲𝘃𝗲𝗿𝘆𝗼𝗻𝗲 𝗯𝗹𝗮𝗺𝗲𝘀 𝗶𝘀 𝘁𝗵𝗲 𝘁𝗼𝗼𝗹.

Wrong tool. Wrong AI. Should have used a different platform.

Almost never the real cause.

In my experience, automation that doesn't stick fails for one of three reasons — and none of them are the tool:

𝗥𝗲𝗮𝘀𝗼𝗻 𝟭: 𝗡𝗼𝗯𝗼𝗱𝘆 𝗼𝘄𝗻𝘀 𝗶𝘁

If responsibility for the automation isn't assigned to a specific person, it belongs to everyone — which means it belongs to no one. It drifts. It gets routed around. It dies quietly.

𝗥𝗲𝗮𝘀𝗼𝗻 𝟮: 𝗧𝗵𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺 𝘄𝗮𝘀𝗻'𝘁 𝗽𝗮𝗶𝗻𝗳𝘂𝗹 𝗲𝗻𝗼𝘂𝗴𝗵

If the process being automated wasn't genuinely hurting anyone, solving it doesn't relieve anyone. Nobody's invested in whether it works because nobody suffered when it didn't.

𝗥𝗲𝗮𝘀𝗼𝗻 𝟯: 𝗜𝘁 𝗮𝘀𝗸𝗲𝗱 𝘁𝗵𝗲 𝘁𝗲𝗮𝗺 𝘁𝗼 𝗰𝗵𝗮𝗻𝗴𝗲 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗲𝘅𝗽𝗹𝗮𝗶𝗻𝗶𝗻𝗴 𝘄𝗵𝘆

A new system imposed on people who don't understand why it exists will be tolerated, not adopted. Tolerance looks like adoption for about 3 weeks. Then old habits return.

𝗧𝗵𝗲 𝗰𝗼𝗺𝗺𝗼𝗻 𝘁𝗵𝗿𝗲𝗮𝗱:

All three are people problems. All three are solvable before you build anything.

The tool is almost never the issue. The conditions around the tool almost always are.

𝗖𝗼𝗺𝗺𝗲𝗻𝘁 "𝗔𝗜" 𝗮𝗻𝗱 𝗜'𝗹𝗹 𝘀𝗲𝗻𝗱 𝘆𝗼𝘂 𝘁𝗵𝗲 𝗳𝗿𝗲𝗲 𝗴𝘂𝗶𝗱𝗲 𝗼𝗻 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻𝘀 𝘆𝗼𝘂𝗿 𝘁𝗲𝗮𝗺 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗮𝗱𝗼𝗽𝘁

[Alex @AgenticSoul]

𝗨𝗻𝗽𝗼𝗽𝘂𝗹𝗮𝗿 𝗼𝗽𝗶𝗻𝗶𝗼𝗻:

"Start small with AI" is the wrong advice for most UK businesses in 2025.

It made sense two years ago when the tools were unreliable and the failure cost was high. That's not the landscape anymore.

Starting small means: low-stakes automation, low-stakes results, low organisational commitment, and a pilot that demonstrates AI "works" without changing anything that matters.

𝗪𝗵𝗮𝘁 𝗜 𝗯𝗲𝗹𝗶𝗲𝘃𝗲 𝗶𝗻𝘀𝘁𝗲𝗮𝗱:

Start focused. Not small — focused.

Find the process where the time loss is most painful, the stakes are genuine but manageable, and success is visible within 30 days. Build that, with the appropriate review layer, and review it honestly at day 30.

That's not "starting big." It's starting in the right place.

The businesses I've seen get the most traction didn't crawl toward AI. They solved one real problem well, saw the result, and built momentum from there.

Crawling doesn't build momentum. Getting one thing genuinely right does.

The "start small" advice keeps businesses permanently in pilot mode. Pilots don't change businesses.

𝗔𝗺 𝗜 𝘄𝗿𝗼𝗻𝗴? 𝗛𝗮𝘀 𝘀𝘁𝗮𝗿𝘁𝗶𝗻𝗴 𝘀𝗺𝗮𝗹𝗹 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝘄𝗼𝗿𝗸𝗲𝗱 𝗳𝗼𝗿 𝘆𝗼𝘂?

[Alex @AgenticSoul]

𝗜𝗳 𝘆𝗼𝘂𝗿 𝗔𝗜 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻 𝗶𝘀𝗻'𝘁 𝘀𝘁𝗶𝗰𝗸𝗶𝗻𝗴, 𝘁𝗵𝗲 𝗳𝗶𝗿𝘀𝘁 𝘁𝗵𝗶𝗻𝗴 𝗲𝘃𝗲𝗿𝘆𝗼𝗻𝗲 𝗯𝗹𝗮𝗺𝗲𝘀 𝗶𝘀 𝘁𝗵𝗲 𝘁𝗼𝗼𝗹.

Wrong tool. Wrong AI. Should have used a different platform.

Almost never the real cause.

In my experience, automation that doesn't stick fails for one of three reasons — and none of them are the tool:

𝗥𝗲𝗮𝘀𝗼𝗻 𝟭: 𝗡𝗼𝗯𝗼𝗱𝘆 𝗼𝘄𝗻𝘀 𝗶𝘁

If responsibility for the automation isn't assigned to a specific person, it belongs to everyone — which means it belongs to no one. It drifts. It gets routed around. It dies quietly.

𝗥𝗲𝗮𝘀𝗼𝗻 𝟮: 𝗧𝗵𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺 𝘄𝗮𝘀𝗻'𝘁 𝗽𝗮𝗶𝗻𝗳𝘂𝗹 𝗲𝗻𝗼𝘂𝗴𝗵

If the process being automated wasn't genuinely hurting anyone, solving it doesn't relieve anyone. Nobody's invested in whether it works because nobody suffered when it didn't.

𝗥𝗲𝗮𝘀𝗼𝗻 𝟯: 𝗜𝘁 𝗮𝘀𝗸𝗲𝗱 𝘁𝗵𝗲 𝘁𝗲𝗮𝗺 𝘁𝗼 𝗰𝗵𝗮𝗻𝗴𝗲 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗲𝘅𝗽𝗹𝗮𝗶𝗻𝗶𝗻𝗴 𝘄𝗵𝘆

A new system imposed on people who don't understand why it exists will be tolerated, not adopted. Tolerance looks like adoption for about 3 weeks. Then old habits return.

𝗧𝗵𝗲 𝗰𝗼𝗺𝗺𝗼𝗻 𝘁𝗵𝗿𝗲𝗮𝗱:

All three are people problems. All three are solvable before you build anything.

The tool is almost never the issue. The conditions around the tool almost always are.

𝗖𝗼𝗺𝗺𝗲𝗻𝘁 "𝗔𝗜" 𝗮𝗻𝗱 𝗜'𝗹𝗹 𝘀𝗲𝗻𝗱 𝘆𝗼𝘂 𝘁𝗵𝗲 𝗳𝗿𝗲𝗲 𝗴𝘂𝗶𝗱𝗲 𝗼𝗻 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻𝘀 𝘆𝗼𝘂𝗿 𝘁𝗲𝗮𝗺 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗮𝗱𝗼𝗽𝘁

[Alex @AgenticSoul]

𝗡𝗶𝗮 𝗪𝗶𝗹𝗹𝗶𝗮𝗺𝘀 𝘄𝗮𝘀 𝗾𝘂𝗼𝘁𝗲𝗱 £𝟴,𝟬𝟬𝟬 𝗯𝘆 𝗵𝗲𝗿 𝗮𝗰𝗰𝗼𝘂𝗻𝘁𝗮𝗻𝘁.

For a system to automate her supplier payment reconciliation.

She built it herself in a week. Total cost: £210 in tooling.

Nia co-founded a Welsh food products brand — 9 staff, 34 active suppliers, 3 retail contracts. Every month her finance manager spent 3 days matching purchase orders to invoices to delivery notes across 34 suppliers who each sent documents in slightly different formats.

The accountant's £8,000 proposal was for custom integration work. Nia took one look at the quote and decided to understand the problem herself before paying for a solution.

𝗪𝗵𝗮𝘁 𝘀𝗵𝗲 𝗯𝘂𝗶𝗹𝘁:

A simple AI document processing flow. Invoices and delivery notes come in via email, get parsed automatically, and are matched against the purchase order log. Exceptions — where numbers don't match — are flagged for human review. Everything clean processes without anyone touching it.

𝗧𝗵𝗲 𝗽𝗮𝗿𝘁 𝘁𝗵𝗮𝘁 𝗳𝗮𝗶𝗹𝗲𝗱 𝗳𝗶𝗿𝘀𝘁:

Two suppliers sent invoices as scanned images rather than PDFs. The text extraction produced garbled figures. Three invoices were flagged as mismatches that weren't actually mismatches.

Nia caught them in review. We added optical character recognition handling for image-based documents and a confidence threshold: any extraction scoring below 85% routes to manual review automatically.

𝗙𝗼𝘂𝗿 𝗺𝗼𝗻𝘁𝗵𝘀 𝗼𝗻:

Reconciliation time: 3 days/month → 4 hours
Finance manager's freed time: directed into supplier relationship management and cost negotiation — work she was trained for and hadn't had bandwidth to do
One renegotiated supplier contract: saving £3,800 annually

The £8,000 system would have done more or less the same thing.

What's a problem in your business where the expensive quote made you find the cheaper answer?

[Alex @AgenticSoul]

𝗛𝗲𝗿𝗲'𝘀 𝘄𝗵𝘆 𝗺𝗼𝘀𝘁 𝗔𝗜 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀 𝗶𝗻 𝗨𝗞 𝗦𝗠𝗕𝘀 𝘀𝘁𝗮𝗹𝗹 𝗮𝘁 𝗺𝗼𝗻𝘁𝗵 𝟯.

Month 1: Exciting. The pilot works. Everyone can see the time saving.

Month 2: Normal. The system is running. It's been delegated to whoever "does the tech stuff."

Month 3: Quiet. Something small broke. Nobody fixed it immediately because it was "mostly working." A workaround appeared. The workaround became the new process. The automation now runs in parallel with the manual version that replaced it.

Month 4: The automation is technically still running. Nobody is sure if it's actually being used.

This isn't a technology problem. It's an ownership problem.

Every automation needs a named owner who: reviews it weekly, is empowered to flag problems, and has a direct line to whoever can fix them.

Without that, "month 3 stall" is almost inevitable. The system doesn't degrade. The attention does.

The fix costs nothing. It's just a calendar reminder and a 15-minute check-in.

𝗦𝗼 𝘄𝗵𝘆 𝗱𝗼𝗲𝘀 𝗻𝗼𝗯𝗼𝗱𝘆 𝘀𝗲𝘁 𝗶𝘁 𝘂𝗽?

[Alex @AgenticSoul]

𝗨𝗻𝗽𝗼𝗽𝘂𝗹𝗮𝗿 𝗼𝗽𝗶𝗻𝗶𝗼𝗻:

"Start small with AI" is the wrong advice for most UK businesses in 2025.

It made sense two years ago when the tools were unreliable and the failure cost was high. That's not the landscape anymore.

Starting small means: low-stakes automation, low-stakes results, low organisational commitment, and a pilot that demonstrates AI "works" without changing anything that matters.

𝗪𝗵𝗮𝘁 𝗜 𝗯𝗲𝗹𝗶𝗲𝘃𝗲 𝗶𝗻𝘀𝘁𝗲𝗮𝗱:

Start focused. Not small — focused.

Find the process where the time loss is most painful, the stakes are genuine but manageable, and success is visible within 30 days. Build that, with the appropriate review layer, and review it honestly at day 30.

That's not "starting big." It's starting in the right place.

The businesses I've seen get the most traction didn't crawl toward AI. They solved one real problem well, saw the result, and built momentum from there.

Crawling doesn't build momentum. Getting one thing genuinely right does.

The "start small" advice keeps businesses permanently in pilot mode. Pilots don't change businesses.

𝗔𝗺 𝗜 𝘄𝗿𝗼𝗻𝗴? 𝗛𝗮𝘀 𝘀𝘁𝗮𝗿𝘁𝗶𝗻𝗴 𝘀𝗺𝗮𝗹𝗹 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝘄𝗼𝗿𝗸𝗲𝗱 𝗳𝗼𝗿 𝘆𝗼𝘂?

[Alex @AgenticSoul]

𝗡𝗶𝗮 𝗪𝗶𝗹𝗹𝗶𝗮𝗺𝘀 𝘄𝗮𝘀 𝗾𝘂𝗼𝘁𝗲𝗱 £𝟴,𝟬𝟬𝟬 𝗯𝘆 𝗵𝗲𝗿 𝗮𝗰𝗰𝗼𝘂𝗻𝘁𝗮𝗻𝘁.

For a system to automate her supplier payment reconciliation.

She built it herself in a week. Total cost: £210 in tooling.

Nia co-founded a Welsh food products brand — 9 staff, 34 active suppliers, 3 retail contracts. Every month her finance manager spent 3 days matching purchase orders to invoices to delivery notes across 34 suppliers who each sent documents in slightly different formats.

The accountant's £8,000 proposal was for custom integration work. Nia took one look at the quote and decided to understand the problem herself before paying for a solution.

𝗪𝗵𝗮𝘁 𝘀𝗵𝗲 𝗯𝘂𝗶𝗹𝘁:

A simple AI document processing flow. Invoices and delivery notes come in via email, get parsed automatically, and are matched against the purchase order log. Exceptions — where numbers don't match — are flagged for human review. Everything clean processes without anyone touching it.

𝗧𝗵𝗲 𝗽𝗮𝗿𝘁 𝘁𝗵𝗮𝘁 𝗳𝗮𝗶𝗹𝗲𝗱 𝗳𝗶𝗿𝘀𝘁:

Two suppliers sent invoices as scanned images rather than PDFs. The text extraction produced garbled figures. Three invoices were flagged as mismatches that weren't actually mismatches.

Nia caught them in review. We added optical character recognition handling for image-based documents and a confidence threshold: any extraction scoring below 85% routes to manual review automatically.

𝗙𝗼𝘂𝗿 𝗺𝗼𝗻𝘁𝗵𝘀 𝗼𝗻:

Reconciliation time: 3 days/month → 4 hours
Finance manager's freed time: directed into supplier relationship management and cost negotiation — work she was trained for and hadn't had bandwidth to do
One renegotiated supplier contract: saving £3,800 annually

The £8,000 system would have done more or less the same thing.

What's a problem in your business where the expensive quote made you find the cheaper answer?

[Alex @AgenticSoul]

𝗧𝗵𝗶𝘀 𝘁𝗮𝗸𝗲𝘀 𝟲𝟬 𝘀𝗲𝗰𝗼𝗻𝗱𝘀.

It'll tell you whether your AI automation is actually saving time or just moving it.

𝗧𝗵𝗲 𝗮𝘂𝗱𝗶𝘁:

Pick the automation you're most proud of.

Now answer: before the automation, how many minutes did the task take per week?

After the automation, how many minutes does a human spend: setting it up, reviewing outputs, fixing errors, and explaining it to colleagues — per week?

Subtract. That's your actual time saving.

𝗪𝗵𝗮𝘁 𝗺𝗼𝘀𝘁 𝗽𝗲𝗼𝗽𝗹𝗲 𝗱𝗶𝘀𝗰𝗼𝘃𝗲𝗿:

The gross saving is real. The net saving is smaller than expected — because the review time wasn't counted.

That's not a failure. It's data. It tells you where to focus the next improvement: usually on reducing error rates so review time shrinks.

𝗪𝗵𝗮𝘁 𝘀𝗼𝗺𝗲 𝗽𝗲𝗼𝗽𝗹𝗲 𝗱𝗶𝘀𝗰𝗼𝘃𝗲𝗿:

Their automation is net negative. The review and maintenance time exceeds what the task used to cost.

This is more common than anyone admits. And it's fixable — but only once you're measuring honestly.

𝗧𝗵𝗲 𝗰𝗼𝗿𝗿𝗲𝗰𝘁 𝗱𝗮𝘁𝗮 𝗽𝗼𝗶𝗻𝘁 𝗶𝘀𝗻'𝘁 "𝗵𝗼𝘄 𝗺𝘂𝗰𝗵 𝘁𝗶𝗺𝗲 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝘀𝗮𝘃𝗲?"

𝗜𝘁'𝘀 "𝗵𝗼𝘄 𝗺𝘂𝗰𝗵 𝘁𝗶𝗺𝗲 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝘀𝗮𝘃𝗲 𝗻𝗲𝘁 𝗼𝗳 𝗮𝗹𝗹 𝘀𝗲𝗰𝗼𝗻𝗱𝗮𝗿𝘆 𝗲𝗳𝗳𝗼𝗿𝘁?"

𝗖𝗼𝗺𝗺𝗲𝗻𝘁 "𝗔𝗜" 𝗮𝗻𝗱 𝗜'𝗹𝗹 𝘀𝗲𝗻𝗱 𝘆𝗼𝘂 𝘁𝗵𝗲 𝗳𝗿𝗲𝗲 𝗴𝘂𝗶𝗱𝗲 𝘄𝗶𝘁𝗵 𝘁𝗵𝗲 𝗳𝘂𝗹𝗹 𝗻𝗲𝘁-𝘀𝗮𝘃𝗶𝗻𝗴 𝗮𝘂𝗱𝗶𝘁 𝘁𝗲𝗺𝗽𝗹𝗮𝘁𝗲

[Alex @AgenticSoul]

𝗛𝗲𝗿𝗲'𝘀 𝘄𝗵𝘆 𝗺𝗼𝘀𝘁 𝗔𝗜 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀 𝗶𝗻 𝗨𝗞 𝗦𝗠𝗕𝘀 𝘀𝘁𝗮𝗹𝗹 𝗮𝘁 𝗺𝗼𝗻𝘁𝗵 𝟯.

Month 1: Exciting. The pilot works. Everyone can see the time saving.

Month 2: Normal. The system is running. It's been delegated to whoever "does the tech stuff."

Month 3: Quiet. Something small broke. Nobody fixed it immediately because it was "mostly working." A workaround appeared. The workaround became the new process. The automation now runs in parallel with the manual version that replaced it.

Month 4: The automation is technically still running. Nobody is sure if it's actually being used.

This isn't a technology problem. It's an ownership problem.

Every automation needs a named owner who: reviews it weekly, is empowered to flag problems, and has a direct line to whoever can fix them.

Without that, "month 3 stall" is almost inevitable. The system doesn't degrade. The attention does.

The fix costs nothing. It's just a calendar reminder and a 15-minute check-in.

𝗦𝗼 𝘄𝗵𝘆 𝗱𝗼𝗲𝘀 𝗻𝗼𝗯𝗼𝗱𝘆 𝘀𝗲𝘁 𝗶𝘁 𝘂𝗽?

[Alex @AgenticSoul]

𝗙𝗼𝗿 𝗮 𝘆𝗲𝗮𝗿 𝗜 𝘂𝗻𝗱𝗲𝗿𝗰𝗵𝗮𝗿𝗴𝗲𝗱. 𝗔𝗻𝗱 𝗜 𝗸𝗻𝗲𝘄 𝗶𝘁.

The mistake wasn't not knowing my value. It was measuring the wrong thing.

I was pricing based on my time. How long a project takes multiplied by a day rate. Which meant that as I got faster — as I built better systems, got more efficient, learned what to look for — my prices should have fallen. By my own logic, being better at my job meant earning less.

𝗧𝗵𝗲 𝗿𝗲𝗳𝗿𝗮𝗺𝗲:

A client's first automation saved them £34,000 in the first year. I charged £3,200 to build it.

That's not a pricing model. That's a donation.

The correct pricing question isn't "how long did this take?" It's "what is this worth to the client?" Those two numbers are completely different. The gap between them is where consultants leave money on the table for years.

𝗛𝗼𝘄 𝗜 𝗽𝗿𝗶𝗰𝗲 𝗻𝗼𝘄:

I estimate the value of the outcome in year one. I price at a fraction of that. I have the conversation about the fraction openly.

It makes proposals easier, not harder. Clients who understand the ROI don't negotiate on price the same way.

𝗜𝗳 𝘆𝗼𝘂'𝗿𝗲 𝗰𝗵𝗮𝗿𝗴𝗶𝗻𝗴 𝗯𝘆 𝘁𝗵𝗲 𝗵𝗼𝘂𝗿 𝗳𝗼𝗿 𝘄𝗼𝗿𝗸 𝘁𝗵𝗮𝘁 𝗿𝗲𝘁𝘂𝗿𝗻𝘀 𝗮 𝗺𝗲𝗮𝘀𝘂𝗿𝗮𝗯𝗹𝗲 𝗥𝗢𝗜, 𝘆𝗼𝘂'𝗿𝗲 𝗹𝗲𝗮𝘃𝗶𝗻𝗴 𝗺𝗼𝗻𝗲𝘆 𝗼𝗻 𝘁𝗵𝗲 𝘁𝗮𝗯𝗹𝗲 𝗲𝘃𝗲𝗿𝘆 𝘁𝗶𝗺𝗲.

[Alex @AgenticSoul]

𝗣𝗮𝘆𝗶𝗻𝗴 𝗮 𝗰𝗼𝗻𝘀𝘂𝗹𝘁𝗮𝗻𝘁 𝘁𝗼 "𝗲𝘅𝗽𝗹𝗼𝗿𝗲 𝗔𝗜 𝗼𝗽𝗽𝗼𝗿𝘁𝘂𝗻𝗶𝘁𝗶𝗲𝘀" 𝗶𝘀 𝗮𝗹𝗺𝗼𝘀𝘁 𝗮𝗹𝘄𝗮𝘆𝘀 𝗮 𝘄𝗮𝘀𝘁𝗲 𝗼𝗳 𝗺𝗼𝗻𝗲𝘆.

Not because consultants are bad. Because "exploring opportunities" produces a report, not a result.

You don't need someone to tell you where AI could theoretically help your business. You know your business. You know where the time goes.

What you need is someone who will build the first automation with you, stay accountable for whether it works, and fix it when it doesn't. That's a fundamentally different engagement than a discovery workshop and a slide deck.

The discovery workshop produces insight. The insight produces a to-do list. The to-do list goes into a folder. The folder goes into a drawer.

The businesses I've seen make the most progress paid for implementation, not exploration. They knew roughly what they wanted to fix, found someone to help them fix it, reviewed the results at 30 days, and made a decision about what to do next.

That cycle repeats. The discovery cycle doesn't.

If a consultant's deliverable is a document rather than a working system, ask yourself whether the document will survive the first week after the engagement ends.

Am I being too blunt? Tell me where the exploration model actually worked for you.

[Alex @AgenticSoul]

𝗧𝗵𝗶𝘀 𝘁𝗮𝗸𝗲𝘀 𝟲𝟬 𝘀𝗲𝗰𝗼𝗻𝗱𝘀.

It'll tell you whether your AI automation is actually saving time or just moving it.

𝗧𝗵𝗲 𝗮𝘂𝗱𝗶𝘁:

Pick the automation you're most proud of.

Now answer: before the automation, how many minutes did the task take per week?

After the automation, how many minutes does a human spend: setting it up, reviewing outputs, fixing errors, and explaining it to colleagues — per week?

Subtract. That's your actual time saving.

𝗪𝗵𝗮𝘁 𝗺𝗼𝘀𝘁 𝗽𝗲𝗼𝗽𝗹𝗲 𝗱𝗶𝘀𝗰𝗼𝘃𝗲𝗿:

The gross saving is real. The net saving is smaller than expected — because the review time wasn't counted.

That's not a failure. It's data. It tells you where to focus the next improvement: usually on reducing error rates so review time shrinks.

𝗪𝗵𝗮𝘁 𝘀𝗼𝗺𝗲 𝗽𝗲𝗼𝗽𝗹𝗲 𝗱𝗶𝘀𝗰𝗼𝘃𝗲𝗿:

Their automation is net negative. The review and maintenance time exceeds what the task used to cost.

This is more common than anyone admits. And it's fixable — but only once you're measuring honestly.

𝗧𝗵𝗲 𝗰𝗼𝗿𝗿𝗲𝗰𝘁 𝗱𝗮𝘁𝗮 𝗽𝗼𝗶𝗻𝘁 𝗶𝘀𝗻'𝘁 "𝗵𝗼𝘄 𝗺𝘂𝗰𝗵 𝘁𝗶𝗺𝗲 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝘀𝗮𝘃𝗲?"

𝗜𝘁'𝘀 "𝗵𝗼𝘄 𝗺𝘂𝗰𝗵 𝘁𝗶𝗺𝗲 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝘀𝗮𝘃𝗲 𝗻𝗲𝘁 𝗼𝗳 𝗮𝗹𝗹 𝘀𝗲𝗰𝗼𝗻𝗱𝗮𝗿𝘆 𝗲𝗳𝗳𝗼𝗿𝘁?"

𝗖𝗼𝗺𝗺𝗲𝗻𝘁 "𝗔𝗜" 𝗮𝗻𝗱 𝗜'𝗹𝗹 𝘀𝗲𝗻𝗱 𝘆𝗼𝘂 𝘁𝗵𝗲 𝗳𝗿𝗲𝗲 𝗴𝘂𝗶𝗱𝗲 𝘄𝗶𝘁𝗵 𝘁𝗵𝗲 𝗳𝘂𝗹𝗹 𝗻𝗲𝘁-𝘀𝗮𝘃𝗶𝗻𝗴 𝗮𝘂𝗱𝗶𝘁 𝘁𝗲𝗺𝗽𝗹𝗮𝘁𝗲

[Alex @AgenticSoul]

𝗙𝗼𝗿 𝗮 𝘆𝗲𝗮𝗿 𝗜 𝘂𝗻𝗱𝗲𝗿𝗰𝗵𝗮𝗿𝗴𝗲𝗱. 𝗔𝗻𝗱 𝗜 𝗸𝗻𝗲𝘄 𝗶𝘁.

The mistake wasn't not knowing my value. It was measuring the wrong thing.

I was pricing based on my time. How long a project takes multiplied by a day rate. Which meant that as I got faster — as I built better systems, got more efficient, learned what to look for — my prices should have fallen. By my own logic, being better at my job meant earning less.

𝗧𝗵𝗲 𝗿𝗲𝗳𝗿𝗮𝗺𝗲:

A client's first automation saved them £34,000 in the first year. I charged £3,200 to build it.

That's not a pricing model. That's a donation.

The correct pricing question isn't "how long did this take?" It's "what is this worth to the client?" Those two numbers are completely different. The gap between them is where consultants leave money on the table for years.

𝗛𝗼𝘄 𝗜 𝗽𝗿𝗶𝗰𝗲 𝗻𝗼𝘄:

I estimate the value of the outcome in year one. I price at a fraction of that. I have the conversation about the fraction openly.

It makes proposals easier, not harder. Clients who understand the ROI don't negotiate on price the same way.

𝗜𝗳 𝘆𝗼𝘂'𝗿𝗲 𝗰𝗵𝗮𝗿𝗴𝗶𝗻𝗴 𝗯𝘆 𝘁𝗵𝗲 𝗵𝗼𝘂𝗿 𝗳𝗼𝗿 𝘄𝗼𝗿𝗸 𝘁𝗵𝗮𝘁 𝗿𝗲𝘁𝘂𝗿𝗻𝘀 𝗮 𝗺𝗲𝗮𝘀𝘂𝗿𝗮𝗯𝗹𝗲 𝗥𝗢𝗜, 𝘆𝗼𝘂'𝗿𝗲 𝗹𝗲𝗮𝘃𝗶𝗻𝗴 𝗺𝗼𝗻𝗲𝘆 𝗼𝗻 𝘁𝗵𝗲 𝘁𝗮𝗯𝗹𝗲 𝗲𝘃𝗲𝗿𝘆 𝘁𝗶𝗺𝗲.

[Alex @AgenticSoul]

𝗦𝘁𝗲𝗳𝗮𝗻 𝗕𝗼𝘄𝗲𝗿 𝗯𝘂𝗶𝗹𝘁 𝘁𝗵𝗲 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻.

His team refused to use it.

Not dramatically. Just quietly, consistently, continued doing it the old way. The system ran. Nobody fed it.

Stefan is a management consultant in Bristol running a 7-person firm. The automation was designed to handle client status updates — pulling from project notes and generating weekly summaries instead of team members writing them manually.

It was technically sound. It worked correctly in every test.

𝗪𝗵𝗮𝘁 𝗵𝗲 𝗺𝗶𝘀𝘀𝗲𝗱:

The weekly status update wasn't just an admin task for his team. It was how they processed their own thinking about a project. Writing it helped them notice problems. It was doing two jobs — producing a document and making them think.

Remove the writing, and you remove the processing. Two consultants independently told Stefan their work felt less clear after the automation was introduced.

𝗛𝗼𝘄 𝘄𝗲 𝗳𝗶𝘅𝗲𝗱 𝗶𝘁:

We didn't remove the writing. We automated everything around it.

The AI pulls the project data and structures the document. The consultant fills in one section: their own read of where the project stands and what they're watching. The rest is generated.

The cognitive work stayed. The clerical work went.

Usage: 100% within a week of the rebuild. No pushback.

𝗧𝗵𝗲 𝗹𝗲𝘀𝘀𝗼𝗻:

Before you automate any task, ask the person doing it: "What does this task do for you beyond the output it produces?" The answer changes what you should automate.

[Alex @AgenticSoul]

𝗣𝗮𝘆𝗶𝗻𝗴 𝗮 𝗰𝗼𝗻𝘀𝘂𝗹𝘁𝗮𝗻𝘁 𝘁𝗼 "𝗲𝘅𝗽𝗹𝗼𝗿𝗲 𝗔𝗜 𝗼𝗽𝗽𝗼𝗿𝘁𝘂𝗻𝗶𝘁𝗶𝗲𝘀" 𝗶𝘀 𝗮𝗹𝗺𝗼𝘀𝘁 𝗮𝗹𝘄𝗮𝘆𝘀 𝗮 𝘄𝗮𝘀𝘁𝗲 𝗼𝗳 𝗺𝗼𝗻𝗲𝘆.

Not because consultants are bad. Because "exploring opportunities" produces a report, not a result.

You don't need someone to tell you where AI could theoretically help your business. You know your business. You know where the time goes.

What you need is someone who will build the first automation with you, stay accountable for whether it works, and fix it when it doesn't. That's a fundamentally different engagement than a discovery workshop and a slide deck.

The discovery workshop produces insight. The insight produces a to-do list. The to-do list goes into a folder. The folder goes into a drawer.

The businesses I've seen make the most progress paid for implementation, not exploration. They knew roughly what they wanted to fix, found someone to help them fix it, reviewed the results at 30 days, and made a decision about what to do next.

That cycle repeats. The discovery cycle doesn't.

If a consultant's deliverable is a document rather than a working system, ask yourself whether the document will survive the first week after the engagement ends.

Am I being too blunt? Tell me where the exploration model actually worked for you.

[Alex @AgenticSoul]

𝗧𝗵𝗲 𝟰 𝗔𝗜 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻𝘀 𝗲𝘃𝗲𝗿𝘆 𝗨𝗞 𝘀𝗲𝗿𝘃𝗶𝗰𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝘀𝗵𝗼𝘂𝗹𝗱 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝗵𝗮𝘃𝗲.

Ranked by the number of times I've watched a business owner's face change when they realise they didn't have it:

𝟭. 𝗖𝗹𝗶𝗲𝗻𝘁 𝗱𝗼𝗰𝘂𝗺𝗲𝗻𝘁 𝗰𝗵𝗮𝘀𝗶𝗻𝗴

The personalised, timed follow-up that goes out automatically when a client hasn't submitted what you need. Every firm I've seen build this recovers 8-15 hours a month and shortens their matter/project cycles by days.

𝟮. 𝗣𝗿𝗲-𝗺𝗲𝗲𝘁𝗶𝗻𝗴 𝗰𝗹𝗶𝗲𝗻𝘁 𝗯𝗿𝗶𝗲𝗳𝗶𝗻𝗴 𝗻𝗼𝘁𝗲

Auto-generated summary of a client's account, recent history, and outstanding actions — sent to the relevant team member 24 hours before any meeting. You walk in informed without spending 45 minutes pulling it together.

𝟯. 𝗡𝗲𝘄 𝗰𝗹𝗶𝗲𝗻𝘁 𝗼𝗻𝗯𝗼𝗮𝗿𝗱𝗶𝗻𝗴 𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲

From signed proposal to all paperwork collected and first deliverable started — without a team member managing each step manually. Every day this takes longer than it should is a day the client doubts the decision they just made.

𝟰. 𝗖𝗹𝗶𝗲𝗻𝘁 𝗿𝗲𝗻𝗲𝘄𝗮𝗹 𝗲𝗮𝗿𝗹𝘆 𝘄𝗮𝗿𝗻𝗶𝗻𝗴

A system that flags any client showing reduced activity, slower response times, or declining engagement — 60 days before their renewal. Not so you can pitch harder. So you can have the right conversation before they've already decided to leave.

𝗔𝗹𝗹 𝗳𝗼𝘂𝗿 𝗲𝘅𝗶𝘀𝘁 𝗶𝗻 𝗺𝗼𝘀𝘁 𝘀𝗲𝗿𝘃𝗶𝗰𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀𝗲𝘀 𝗮𝘀 𝗺𝗮𝗻𝘂𝗮𝗹 𝗽𝗿𝗼𝗰𝗲𝘀𝘀𝗲𝘀 𝘁𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻 𝗶𝗻𝗰𝗼𝗻𝘀𝗶𝘀𝘁𝗲𝗻𝘁𝗹𝘆.

𝗔𝗜 𝗺𝗮𝗸𝗲𝘀 𝘁𝗵𝗲𝗺 𝗵𝗮𝗽𝗽𝗲𝗻 𝗲𝘃𝗲𝗿𝘆 𝘁𝗶𝗺𝗲.

[Alex @AgenticSoul]

𝗛𝗼𝘁 𝘁𝗮𝗸𝗲:

The businesses winning with AI in the UK right now aren't the ones you'd expect.

Not the tech companies. Not the VC-backed scale-ups with dedicated AI teams.

It's the 10-person professional services firm that automated its onboarding.
The sole trader accountant who automated client prep.
The 20-person manufacturer who automated their quoting.

They're winning not because they have more resources — because they have fewer. Fewer people means every hour lost to admin is more proportionally damaging. So the incentive to fix it was always higher.

The large organisations are still running committees to "explore AI strategy." The small ones are solving actual problems in actual weeks.

Smaller businesses move faster because the person who feels the pain is also the person who can authorise the fix. There's no procurement process between the problem and the solution.

I've worked with both. The 8-person firm that moved in 6 weeks and the 200-person firm that's still in discovery after 8 months. Same technology available to both.

The competitive advantage in AI right now belongs to whoever decides first.

Am I wrong about this? Genuinely interested in the counterargument.

𝗖𝗼𝗺𝗺𝗲𝗻𝘁 "𝗔𝗜" 𝗮𝗻𝗱 𝗜'𝗹𝗹 𝘀𝗲𝗻𝗱 𝘆𝗼𝘂 𝘁𝗵𝗲 𝗳𝗿𝗲𝗲 𝗴𝘂𝗶𝗱𝗲 𝘁𝗼 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮𝗻 𝗔𝗜-𝗽𝗼𝘄𝗲𝗿𝗲𝗱 𝗦𝗠𝗕 𝗶𝗻 𝗮𝗻𝘆 𝘀𝗶𝘇𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀

[Alex @AgenticSoul]

𝗦𝘁𝗲𝗳𝗮𝗻 𝗕𝗼𝘄𝗲𝗿 𝗯𝘂𝗶𝗹𝘁 𝘁𝗵𝗲 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻.

His team refused to use it.

Not dramatically. Just quietly, consistently, continued doing it the old way. The system ran. Nobody fed it.

Stefan is a management consultant in Bristol running a 7-person firm. The automation was designed to handle client status updates — pulling from project notes and generating weekly summaries instead of team members writing them manually.

It was technically sound. It worked correctly in every test.

𝗪𝗵𝗮𝘁 𝗵𝗲 𝗺𝗶𝘀𝘀𝗲𝗱:

The weekly status update wasn't just an admin task for his team. It was how they processed their own thinking about a project. Writing it helped them notice problems. It was doing two jobs — producing a document and making them think.

Remove the writing, and you remove the processing. Two consultants independently told Stefan their work felt less clear after the automation was introduced.

𝗛𝗼𝘄 𝘄𝗲 𝗳𝗶𝘅𝗲𝗱 𝗶𝘁:

We didn't remove the writing. We automated everything around it.

The AI pulls the project data and structures the document. The consultant fills in one section: their own read of where the project stands and what they're watching. The rest is generated.

The cognitive work stayed. The clerical work went.

Usage: 100% within a week of the rebuild. No pushback.

𝗧𝗵𝗲 𝗹𝗲𝘀𝘀𝗼𝗻:

Before you automate any task, ask the person doing it: "What does this task do for you beyond the output it produces?" The answer changes what you should automate.

[Alex @AgenticSoul]

𝗧𝗵𝗲 𝟰 𝗔𝗜 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻𝘀 𝗲𝘃𝗲𝗿𝘆 𝗨𝗞 𝘀𝗲𝗿𝘃𝗶𝗰𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝘀𝗵𝗼𝘂𝗹𝗱 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝗵𝗮𝘃𝗲.

Ranked by the number of times I've watched a business owner's face change when they realise they didn't have it:

𝟭. 𝗖𝗹𝗶𝗲𝗻𝘁 𝗱𝗼𝗰𝘂𝗺𝗲𝗻𝘁 𝗰𝗵𝗮𝘀𝗶𝗻𝗴

The personalised, timed follow-up that goes out automatically when a client hasn't submitted what you need. Every firm I've seen build this recovers 8-15 hours a month and shortens their matter/project cycles by days.

𝟮. 𝗣𝗿𝗲-𝗺𝗲𝗲𝘁𝗶𝗻𝗴 𝗰𝗹𝗶𝗲𝗻𝘁 𝗯𝗿𝗶𝗲𝗳𝗶𝗻𝗴 𝗻𝗼𝘁𝗲

Auto-generated summary of a client's account, recent history, and outstanding actions — sent to the relevant team member 24 hours before any meeting. You walk in informed without spending 45 minutes pulling it together.

𝟯. 𝗡𝗲𝘄 𝗰𝗹𝗶𝗲𝗻𝘁 𝗼𝗻𝗯𝗼𝗮𝗿𝗱𝗶𝗻𝗴 𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲

From signed proposal to all paperwork collected and first deliverable started — without a team member managing each step manually. Every day this takes longer than it should is a day the client doubts the decision they just made.

𝟰. 𝗖𝗹𝗶𝗲𝗻𝘁 𝗿𝗲𝗻𝗲𝘄𝗮𝗹 𝗲𝗮𝗿𝗹𝘆 𝘄𝗮𝗿𝗻𝗶𝗻𝗴

A system that flags any client showing reduced activity, slower response times, or declining engagement — 60 days before their renewal. Not so you can pitch harder. So you can have the right conversation before they've already decided to leave.

𝗔𝗹𝗹 𝗳𝗼𝘂𝗿 𝗲𝘅𝗶𝘀𝘁 𝗶𝗻 𝗺𝗼𝘀𝘁 𝘀𝗲𝗿𝘃𝗶𝗰𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀𝗲𝘀 𝗮𝘀 𝗺𝗮𝗻𝘂𝗮𝗹 𝗽𝗿𝗼𝗰𝗲𝘀𝘀𝗲𝘀 𝘁𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻 𝗶𝗻𝗰𝗼𝗻𝘀𝗶𝘀𝘁𝗲𝗻𝘁𝗹𝘆.

𝗔𝗜 𝗺𝗮𝗸𝗲𝘀 𝘁𝗵𝗲𝗺 𝗵𝗮𝗽𝗽𝗲𝗻 𝗲𝘃𝗲𝗿𝘆 𝘁𝗶𝗺𝗲.

[Alex @AgenticSoul]

"𝗔𝗜 𝘄𝗶𝗹𝗹 𝗿𝗲𝗽𝗹𝗮𝗰𝗲 𝘆𝗼𝘂𝗿 𝘀𝘁𝗮𝗳𝗳."

This is the headline that gets clicks. It's not what's actually happening in UK SMBs.

Here's what's actually happening:

𝗪𝗵𝗮𝘁 𝗔𝗜 𝗶𝘀 𝗿𝗲𝗽𝗹𝗮𝗰𝗶𝗻𝗴: the parts of your staff's jobs that nobody wants to do. The data entry. The chasing. The same-question emails. The report formatting.

𝗪𝗵𝗮𝘁 𝘆𝗼𝘂𝗿 𝘀𝘁𝗮𝗳𝗳 𝗰𝗮𝗻 𝗱𝗼 𝘄𝗵𝗲𝗻 𝘁𝗵𝗮𝘁 𝗴𝗼𝗲𝘀: the work they were actually hired for.

Every business I've worked with where AI has freed significant staff time has had the same conversation: morale went up, not down. People are doing more of the work that requires their brains. They're doing less of the work that makes them feel like expensive printers.

The business owners worried about replacing people with AI are usually in businesses where AI will do the admin and the people will do the actual job. Which is precisely what the people wanted.

The thing to be afraid of isn't AI taking jobs. It's your competitors using AI to do more with the same headcount while you're still manually producing the same documents you produced in 2019.

𝗧𝗵𝗮𝘁'𝘀 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝗿𝗶𝘀𝗸.