Top 10 AI Success Factors
AI strategy is far more important than tactics.
I’m getting weary with the pace and volume of artificial-intelligence buzzwords and fads. One minute everyone is talking about multi-modal AI, then it is prompt engineering, then agentic AI and so on. Are you using the latest & greatest LLM ?? How many billion (or trillion) parameters does the model have ??
This is all interesting, but how AI is applied is way more important than the mechanics within.
In this post, I go back to basics by compiling my top 10 success factors for AI from a business strategy viewpoint (in no particular order).
#1 Making it too complicated
It is a common misconception that any AI solution has to be mind-blowingly awesome.
Simple, quick and easy solutions (in line with a solid longer-term strategy) are frequently much better than complex ones. Sometimes the mundane has most value.
For example: a lightweight predictive machine learning model might have more business value than a computationally & financially expensive generative AI model.
Often, the best solution is much simpler than you might think. Rapid & incremental improvements over time beats a single big-bang approach.
#2 Thinking AI is the silver bullet
Don’t think that AI will automatically solve all your problems. The challenge is usually in moving from a temporary AI wow factor to something that is genuinely useful & valuable, & is therefore enduring.
Process and people are often much more important than technology alone. So:
I’d choose good people who use simple tools very well, rather than average folk with cutting-edge tools.
As I detail in #3, start with the business, not with the technology.
#3 Not having a strategy
An AI strategy need not be complicated. In fact, the more concise, the better. However, it does need to be carefully considered with clear objectives and measures of success.
Start with the business issues such as challenges, constraints and opportunities. Then work backwards from there. For example, if:
- you make widgets
⇒ inspecting widgets as they come out of the widget-maker is slow, expensive and tedious
⇒ a custom computer vision system could detect good & bad widgets. Your AI strategy might include widget quality inspection as a key area, with a target of saving x% or reducing defects by y% - you are an architect
⇒ project concept design takes 10% of effort and contract documentation takes 90%
⇒ a generative AI conceptual design tool might have a wow factor, but making the CD part more efficient should be the priority
There is obviously a human aspect here- it’s not all about the numbers:
- can you eliminate a dull task that no-one likes doing ?
- are people with certain roles/skills stretched so thin that they burn out?
helping these people out might be a sound strategy
It is unnecessary to engage a fancy & expensive business consultant for this task, since you surely know your own people & business better than anyone else. You might need some expert advice to know what is & isn’t possible or to make it happen.
#4 Giving the wrong person (or people) the job
Brainy & well-intentioned, but very technical AI or ML people may not be able to see the forest for the trees (branches →leaves→stomata……) For these people, AI is not a means to an end, but the end itself.
A balance is required: enough technical knowledge to make an informed decision, but not so much that it clouds everything else.
Equally, committees are great at discussion, but poor at decision making and even worse at execution. They tend to dilute great ideas & elevate mediocre ones.
So: decisiveness and action is better than watered-down ideas or nothing at all.
Find good people who get things done & give them the latitude to make decisions. Accept failures alongside successes.
#5 Starting with the technical details
Technical folk (I’m one) are inclined to get into irrelevant details way too early. Ask for a metaphorical AI ‘game plan’ and you’ll get a lengthy technical specification, or load of business-consultant-jargon or AI-lingo.
This AI game plan should just explain who is going to pick up the ball and the direction they will run.
The details matter at some point in time, but only need to be considered in meeting a specific goal or target.
#6 Believing your own bullshit
All organisations need to promote themselves and every board wants a good-news story. AI suits this well, with themes of innovation, advancement or competitive advantage.
However, a cosmetic AI veneer will be counterproductive in the longer term. Perceptions of superficiality are difficult to shake & colleagues & customers will dismiss future initiatives to the same category.
To gain credibility, trust & future support, the focus needs to be on genuine & sustainable benefit.
#7 Believing someone else’s bullshit
Software vendors are obviously motivated by making a sale, so it is unsurprising that an AI sales pitch will usually contain wildly optimistic (if not patently false) claims. Or things that have nothing to do with AI are branded as such.
Your competitors will claim to have far superior capabilities and it is easy to think that you have or are falling behind (and that buying that AI subscription will instantly fix the problem)
Big business consultancies are good at selling consulting hours & producing glossy reports full of impressive but meaningless graphics & jargon. The usual outcome of these reports is that you need to pay the same consultant 50x more for the next project phase.
So: bring it back to business fundamentals such as saving time & money, improving performance, freeing up people for higher value & more fulfilling work, or offering a better product/service.
#8 Having unrealistic expectations
Tied in to #6 and #7, it is easy to get caught up in what you think others are doing, or believing your own hyperbole or the marketing guff.
If you are just buying an off-the-shelf AI solution, then your competitors can do exactly the same so there is unlikely to be a significant benefit or business differentiator.
The old adage ‘garbage in, garbage out’ in particularly relevant in a machine learning or AI context- you can’t predict how many pies will sell tomorrow if you don’t know how many sold last year.
The tool itself is less important than how & where it is applied. This doesn’t come easily in a business environment- the ‘soft’ aspects such as communication, convincing others, explaining it, supporting it, promoting it are often more challenging than the ‘hard’ technical AI stuff.
#9 Forgetting the ‘why’
Elsewhere in this blog, I have covered the tendency for excitement in new technology to overshadow rational decision-making. The result is the use of technology just ‘because you can‘.
Tangible, realistic and measurable goals are essential:
Why are we doing this? What improvements in efficiency are we aiming for? What cost reductions do we expect? What risks will be reduced? How will we measure our success (or failure)
Despite the AI hype, the ‘why’ must always be the driver for the ‘how’
#10 Just adding more stuff (not replacing)
Organisational change is much harder than technology change. You have to deal with departments & people who have built their entire existence or career on a particular application, system or process. I’ve worked in many organisations where a particular way of doing things is so entrenched, colleagues will acknowledge ‘it doesn’t make sense, but that’s the way we’ve always done it’
The tendency is just to layer new on top of the old stuff. But you need to be prepared to replace things that have may have worked for years.
For example, I’ve been working with an organisation who have core technology systems that are 30+ years old. Sure, we could tack some AI onto it- but it doesn’t make any sense. The real answer is to replace some of these older systems and for the business processes & human skills to change.
Wrap-up
This list is by no means exhaustive and there are many other factors that should be considered. I have covered some of these topics in more detail elsewhere on this blog, so I’d encourage you to browse through these posts
If you or your organisation need help in any of these areas, please feel free to get in touch. An independent view and a fresh set of eyes can see the forest as well as the trees.
This post is an adaptation of an earlier post 'https://hannellbim.com/2018/09/07/digital-engineering-top-10-mistakes/'
I've kept the same headings, but adjusted it to an AI context.
'the more things change, the more they stay the same'
Andrew Hannell: POV 