Some uncomfortable facts about failures

In the last 18 months, AI has matured sufficiently for real business adoption. However examples of low to no impact are all too frequent.

RAND - the policy and research organisation - puts the AI project failure rate above 80%, roughly twice the rate of conventional IT. They also found that miscommunication about a project’s purpose caused more failures than the technology itself.

The 4 most common causes were:

1. the project was interesting rather than expensive

2. nobody agreed what the right outcomes looked like

3. it sat outside the tools people were using, and

4. nobody owned it

Almost every business leader they spoke to had the same story; somebody had trialed a tool, the demo looked impressive, a few people tried it for a couple of weeks, then with no formal decision being taken they just stopped.

The pilot-to-production gap

MIT’s Project NANDA found 95% of organisations investing in generative AI saw no measurable return. If you find yourself in one of these unfortunate majorities, we here at 48X are seeing things differently.

One of our customers, an AI security platform, is saving £270,000 per annum in the engineering team alone, with support tickets down 60%. And that's from some AI enabled workflows that took just 2 days to insert. Straight into production.

48X has plenty more where this came from, across the GTM function, through forecasting and finance and across the back-office.

Is this you ?

• An AI tool has been trialed and gone nowhere

• Colleagues regularly asking for information that already exists

• If asked, you'd have to think about who owns AI

A quick test to identify candidate workflows

Score a bunch of proposed workflows 1-5 for each of these questions:

1. How often is the task repeated (1 is rarely, 5 is often)

2. What does it cost the business

3. Could your team unanimously agree on what good looks like, and

4. Does the data live in email, documents or a system

If any of your workflows score above 12, they are strong candidates for a short discovery call.

More research that confirms wasted time

Quickbase found 59% of employees spend 11 or more hours a week chasing information across systems. In a 20 person business, if just 10 of those people are doing this, that's roughly 3 very unproductive full-time staff. At $40,000 each, that's $120,000 off the bottom line.

Run the numbers yourself.

Several questions you should be asking your AI lead

1. Which workflow are you doing first, and by when ?

2. What does good look like, and whose agreed that ?

3. Where do we consume the output ?

4. Can I see the source of every answer, if I need to check ?

5. Whose accountable, on both the technical side and the business ?

6. What happens to my data, and can you show me an audit trail ?

7. What's the mini-business case ?

What will 48X do for you

• We will install AI into your workflows, one at a time, based on tools and process in use.

• Whatever tools you're missing we can usually backfill from our platform, no more licence costs.

• Your AI workflows are protected by EarlyCore, our agentic security partner. Your data stays safe.

• Your portal provide analysis of each user's AI usage across newly supercharged workflows, alongside insights and recommendations. This can also be switched on for users outside the 48X workflows, giving you a detailed view across the business about how optimised your AI usage is, where the gaps are and who could be coaching who.

We can start with a 20 minute discovery call, and thereafter a longer planning call.

We spend 1-2 days implementing and testing, aiming at go-live in less than 48 hours.

Doesn't cost much either, From £5,000 to £30,000 a year, plus a small implementation fee.

Call us.