Skip to content 
With AI (or the very least LLMs) now firmly entrenched in almost all aspects of life, both professional and personal, it’s not a matter of which tools are the best, but rather how to use those that best fit your circumstances.
As with all technology, there is a learning curve involved, and by learning how to use such tools most effectively, you are going to be better able to extract the enormous potential that they hold.
But as is often the case with new tech, the temptation is to overcomplicate things to the point where they can rapidly become a drain on company resources, limiting their effectiveness and adding unnecessary complications that your employees have to spend more time and resources on figuring out how to overcome them.
Fortunately, with a bit of oversight and a common-sense approach, it’s possible to ensure the converse of all of that and end up in a situation that the hyperscalers that created these tools envisioned, whereby they become a means to an end rather than an end to themselves.
Why AI Tools Often Become Overly Complex
As with any new technology that has seemingly exploded onto the scene, it takes time for it to bed into existing workflows and for employees and management teams to fully get to grips with it.
When it comes to AI in particular, this poses some issues that are slightly different from other new introductions, and often come about thanks to the sort of unknown nature of it. When used correctly, businesses can achieve amazing efficiency and productivity boosts, but when not implemented adequately, it will result in AI bloat and pushback from those who are being forced to use it more and more.
Another part of the issue is that complexities don’t often come right away, but creep in over time as you scale up systems and add in the latest and greatest features usually touted by the enormous companies behind them.
Feature Creep And Overengineering
In essence, feature creep occurs when new capabilities are added without considering whether users actually need or understand them. This can manifest in a number of ways, all of which often add unnecessarily new features that only serve to complicate existing processes.
- Adding options to satisfy edge cases: Adding in new options to cover all bases may seem like a great idea from the perspective of the company behind the AI, but for the average user, it simply places an even greater burden to enforce new elements that they might not need and spend more time training staff.
- Designing for hypothetical users instead of real ones: Although this isn’t always the case, it can appear that new additions to a solution are designed for a hypothetical customer rather than those who are actually using and operating these systems right now.
- Confusing power with usability: This is where many businesses just starting to include AI into their workflows trip up. When you conflate additional power with usability, you are simply going to end up paying more and only eking out diminishing returns.
Poor Alignment Between AI and Human Workflows
If you are currently an employee of a company that is trying to force new AI solutions into its operations, you will know all too well this issue. While there might be a bigger picture at play behind the scenes, when management tries to push in new AI tech too soon, or before they’ve fully taught their teams how to make the best use of it, it can seriously hinder productivity, causing the inverse of what you actually want to happen.
How To Overcome Such Drawbacks
It might seem that from reading to this point in this post, we are particularly anti-AI. This is certainly not the case, and with the correct execution and a real understanding of why and how you will use the technology, you can avoid many of the main drawbacks.
Start With Clear Intent, Not Shiny Features
It’s tempting to get suckered into the sales talk and promise of a future where AI does all of the hard work, leaving humans to focus on the things that count. If you find yourself pulled towards the lovely talk, you’re not alone.
This kind of sales patter has enamoured almost all executives and Wall Street professionals and has seen the stock market rise over the past year to an incredible degree. To achieve the best results, you will want to take a step back and think about why you need the tools.
Before adopting any new AI tool, force yourself to complete the sentence: ‘This tool will reliably help me do X better or faster.’ If you can’t fill in the blank in under ten seconds, walk away.”
Key takeaway: Go into the selection process with your eyes open to what you need, not want.
Favor Depth Over Breadth
There are a myriad of different LLMs on the market, and that’s only counting the hyperscalers like Google, OpenAI, and Grok that actually create the backbone of it. If you are looking at the tools that tap into these models and provide additional automation to various other tasks, you are looking at hundreds to choose from.
This creates a situation where it’s easy to keep adding new platforms to your stack to try and accomplish different tasks. Instead, look for fewer platforms that can achieve the outcomes you desire under one roof. Not only will this be easier to add to current workflows, but it also makes life easier when requiring the more mundane things like support, etc.
Key takeaway: It’s typically better to choose one or two platforms instead of adding more and more to your stack.
Build Lightweight Workflows
Even if you follow the previous two tips and end up with a leaner stack that ought to accomplish your goals, it’s still up to you to ensure that you create lightweight workflows that live within no more than one or two applications.
Key takeaway: A lighter touch will often yield far better results than if you lean too heavily into multiple programs.
Tool Selection Framework: Key Effectiveness Metrics To Focus On
If you want to be totally sure that you end up with the right tool for your needs, you need to know which metrics will be most valuable.
| Metric | What it measures | Why it matters |
| Accuracy | Correct outputs | Accuracy is crucial for all business operations, and if a tool is making too many mistakes, it will reduce morale and add additional work for your staff. |
| Latency | Response time | One of the reasons you might choose to implement AI into operations is to speed things up. If it takes too long to crunch the numbers, so to speak, you may waste too much time. |
| Adoption | User usage rates | Although training your staff to use new tools is the responsibility of the management team, if they choose to avoid it, it could be something you need to address. |
| Error rate | Failures or hallucinations | AI LLMs are well-documented for hallucinating with their answers. Although much less than in the past, it is still something you need to keep an eye on. |
| Maintenance effort | Time to manage | If you are spending as much time putting out the fires a platform is causing as it is saving your business, you might have chosen the wrong platform or application. |
Automate Carefully…Not Everywhere
It seems that for many top-level executives, the name of the game is automation. The more they’re able to automate, the more money and resources they can save on staff and therefore boost productivity.
It’s certainly the case that AI is almost perfectly designed to automate the kinds of mundane tasks that take up a lot of human time, but trying to automate everything can cause unintended effects that might coalesce into issues far more serious than you anticipate (think making redundant members of staff only to find that you cut too fast, too soon).
Keep Humans In The Loop
The fear of AI taking over the average Joe’s job is definitely real, and according to some data already happening at a rapid clip. But to do so in a haphazard manner will decimate your workforce morale, causing a drop in productivity across the board, and leave you open to some serious problems later down the line.
Maintain AI Tools With Regular Simplification Reviews
It’s not only about choosing the right tools and slitting them perfectly within your organization, but an ongoing effort to ensure that they’re still working as intended, or if you might have outgrown or overspent and need to upgrade or scale back, respectively.
- Conduct regular audits: Just as you ought to be doing vis-à-vis most aspects of your workflows, conducting regular audits can give you a much clearer picture of how your employees are using the tools, if they are meeting expectations, or if there are any holes in the implementation that need plugging.
- Remove before you add: The golden rule, if you’re serious about reducing complexity, is to look at what you can remove before you consider adding more. It’s not always easy to do so, but spending a bit of time looking at things from the outside and how you might use something can give you more clarity and see if it’s possible to remove and therefore save resources, rather than having to add even more expenses to your income statement.
AI is absolutely one of the more revolutionary bits of technology that has graced our timeline, and in many ways can be considered a paradigm shift in how we work. But it’s not entirely infallible, and without the right way of looking at it, you can end up spending more money without seeing the cost savings and efficiency boosts you might have liked to.
