Month: January 2026

Posted on

How To Help Managers Make Decisions

2026-01-08_crawling.png

Why is it so hard to get approval for sustainbility projects even when we know how important it for the planet that we take action?

We need to understand how managers make decisions:

Importance

Sustainability may be important to us but it’s a small part of what managers have to deal with, perhaps 3% to 5% of their load.

Impact

Projects often have a low ROI or high payback. The numbers are often too small to make a difference.

Apathy

There’s no urgency to do more than the minumum. As long as you comply with regulations everything’s fine.

Risk

Taking action is risky. There’s no way to AI your way to a safe and no-risk business case. You’re asking people to make a bet – and most of are much more afraid of making a wrong call than getting it right.

Importance, impact, apathy and risk. Just some of the factors you’re dealing with when trying to get a project approved.

The bureacratic response is to create an approvals process. A pipe to crawl through. Except no one tells you that while you can enter the pipe easily – go and build a business case – it narrows more and more until there’s no way to move forward.

So what do you do in this situation.

Recognise that you’re asking people to place a bet, not giving them certainty.

There’s a risk/reward assymetry.

So, you have to figure out how to make your project the logical or obvious choice.

Give the client the ability to walk back. Give them options to bail out.

Work out what they really want to do and design your project so what it does makes it easier for them to get what they want.

Do all that – and you can turn that pipe you’re crawling through into a slide that you can slide down.

Posted on

A Useful List Of Audit Tests For Data Quality

2026-01-07_audit-checks.png

It’s the time of year when many companies start wrapping up their numbers ready for reporting and audit.

Here are six checks that we find useful when processing data for our clients.

1. BAD DATA

If you’re collating data in spreadsheets there’s a decent chance that hidden in there are /As and that have been missed. Make sure all errors are out.

2. CROSS CHECKS

Build in cross checks into every sheet. Do the summary totals match the detail totals?

3. YEAR ON YEAR VARIANCE

Check for variance at multiple levels.

At a total level, do the numbers look in line? What if you drill down a bit more?

What’s your threshold. Any variance over 20% will need a closer look.

4. COMPLETENESS

High level variances often occur because data is missing.

Do a completeness check – is everything there that you expect or are there holes?

And if there are holes, can you get the missing data in time or do you need to patch with estimates?

5. OUTLIERS

Now drill into the detail.

Say you have 12 numbers for the year for a particular source, like electricity usage.

Do they look reasonable, in line with seasonal trends, or are there outliers that need explaining?

6. LIKE FOR LIKE VARIANCE

Next look at like for like detail.

Compare each source this year to the same period last year.

Your auditors won’t tell you the exact tests they’re using or the thresholds they use to identify anomalies.

But if you do these checks you’ll be in a good position for an audit – and hopefully the observations and improvement points will be minor.

One final thing – where should you do these checks?

They need to be done BEFORE you load data into a system. It’s the GIGO principle – get the dataset ready and clean before you try and feed it into the next stage of a process.

Get in touch if you’d like to see how we do this in practice.

Posted on

How Are You Using, Or Not Using, AI

2026-01-06_ai-systems.png

New technologies enable us to do things that weren’t possible before.

I recently read “The Notebook: A history of thinking on paper” by Roland Allen.

It argues that there are types of thinking that we just can’t do without a technology like paper.

Complex maths. Detailed arguments. Scientific research.

So what does AI technology help us do that we couldn’t do earlier?

It takes a lot of time and effort to build deterministic systems – something that provides a predictable, reliable and repeatable output.

That’s because I have limited time, knowledge and capacity – and I have to work within those constraints.

AI has sped that up. I’ve rebuilt in two days something that took me six months to create the first time around.

The problem with AI is that it is stochastic – and works on probabilities rather than certainties.

As a result it mimics creative, expressive and structured work – creating text, images, videos – or ingesting your mess of meeting notes and coming up with a passable summary.

Using this output directly is problematic. It comes across as false, as if you haven’t put the work in, like you’re passing off something else as your own.

But its unpredictable output is fine in situations like a chatbot that helps you interrogate a knowledge base, or a shopping assistant.

Here’s where I’ve leaned into using AI

  1. Generating code
  2. Fixing code
  3. Strategy and planning
  4. Research
  5. Drafts for compliance and boilerplate copy
  6. Text to voice
  7. Image animation
  8. Avatars that narrate text

And where I wouldn’t use it (yet).

  1. Writing posts / papers / books
  2. Outreach
  3. Decision support (without careful study)

Like most people, I’m still working through how to integrate these new technologies into my ways of working.

But I think it’s safe to say that using it will make it possible for you to do things that were pretty much impossible before.

How are you choosing to use or not use AI?