Using AI for Workflow Optimization and Software Development

Written by: Danny Kleine
Solutions.io

For quite a while now, AI has become all the hype. It is everywhere and everyone seems to be using it. There is a lot to be said and the internet is filled with ideas and articles about AI. I decided to pile on a little with what we have seen and experienced in the last months, and where I think all this will lead.

What Can You Do with AI?

There are many applications. Too many to count almost. I have found that as soon as you start using AI tools, you'll start thinking about the possibilities in a different way and it's easier to incorporate it in your workflows.

The other day I cooked my first AI-generated recipe. Basically, I took a picture of my fridge and told the robots to tell me what I could do with my leftovers. Besides some hallucinations about what vegetables were in my fridge, I got decent recommendations.

The outcomes vary. Not everything is useful or helpful. Overall, I'd say start simple and slowly build out your tools. Don't try to implement big workflows without many review steps or it will likely fail.

Now let's take a look at some AI implementations. This is not an exhaustive list, just some that we found particularly helpful.

Coding

For us, the most natural starting point. I tried many different tools and so far AI seems to be most suited as an assistant that can help you perform various tasks. Writing tests or reviewing existing code works really well. Making small tools and helpers is also easy enough.

Writing entire software with tools like Loveable? In theory it works. You'll get something that looks decent and might do the job. Things get more difficult if you have to get consistent output and want to stop the tools from making fundamental changes. If you're not aware of the changes being made constantly, you might lose it all (read more here).

Data Analysis

Data-related tasks are much easier now that you can use AI. Web scrapers are made in minutes. Crawlers and parsers to and from different file formats are easily made. The AI can recognize patterns and fill data accordingly. Just be aware it does not alter your data.

Translation and Web Work

Heavily debated is translation. While of course not as good as getting an actual translation done, it can save a lot of money to automatically translate less critical business information. You set up a glossary with important translations and let the AI look at context and do the heavy lifting. Automatic workflows can translate new content as soon as it comes up.

A while back we translated one of our internal platforms to Ukrainian and after letting our lead developer from there check the language, the verdict was "meh, it's ok".

What to Keep in Mind When Using AI

If you want to implement AI in your organization, there are a few important things to keep in mind.

What Tools Are in Use in Your Team?

Many people use AI tools independently of their jobs. This raises the question: do you want to limit usage or centralize usage for your organization?

If everyone uses their own tools, you cannot control where data might land. For example, if you use the free version of GitHub Copilot to write code, the code will be used for their models. This could mean your proprietary code or business logic lands in AI models. If you use team plans, you can often control if and how data should be shared.

This all requires some thinking beforehand, and probably the creation of company guidelines on the use of AI.

Who Gets and Owns Your Data?

When you use AI tools, you both share and receive a large amount of data. You might be sharing sensitive information in your prompts, and this information might be published.

At the same time, many companies are trying to use your input and data to train their own models. A notable example this year was the file transfer service WeTransfer, who tried using shared data to train their models. They quickly had to backtrack after people called them out on this (read more here).

How Will AI (Pricing) Develop?

AI tools are rapidly developing at the moment. There are of course the main players like Claude and OpenAI who take up most of the attention. But at the same time, there are many tailored startups and specialized tools working to develop their own solutions. Finding the correct tool for the job can be challenging.

Fortunately, most of these tools are very affordable at the moment. For 20-50€ per month, you can have a basic tool to experiment with.

At the same time, we already see bigger and more expensive plans forming. I expect that AI tools will become much more expensive as they mature and start providing more predictable value to their users. In fact, it's not even a prediction anymore. OpenAI recently announced new features are locked to the $200/month plan (source).

Guidelines and Conclusion

I found that AI usage involves a lot of sharing with team members. Sharing what is worked on and how it is achieved (or improved) helps to get better outcomes and better structures to work with.

Some guidelines on how to implement and use AI in your organization are good to have. Consider the points written above and decide what is most important for your team. Different teams can have different priorities, so it's good to come up with a solid strategy.

Want to discuss how AI might fit into your workflows? Get in touch with us.