5 Tips on Starting and Managing AI projects (for Business Stakeholders)

With market getting excited in trying out AI and key decision makers willing to make the bet, the onus has now shifted. It’s up to the data scientists, data engineers and architects to realise the true potential of every business. Both business heads and tech heads identified a common ground to work on and we see solutions taking shape.

Although different sectors have different data, workflows and challenges, a few key tenets hold true for all. Below are a few learnings and tips on starting-managing AI projects that I recommend to my clients. It applies to any business stakeholders:

1. Get Started ‘First’:

I am sure that when people were working on making cell phones in the 80s, somebody would have imagined a palm size piece of jewelry with clear display, working with touch and voice. But note that they didn’t scrap the ongoing development of those big phones that took huge power, looked ugly, had limited functionality and yes… cost a bomb. With all it’s demerits, it still sold. The reason it sold was because it solved a problem and it was better than not having one at all. I feel the same applies to AI/ML. If you see that you get your RoI, get it done. Don’t obsess over that perfect piece.

2. Phase it Out:

Experienced business managers know that new things keep coming in the market. Some may work really well for you while some may not impress with results. There are chances that some might not work at all. And even with those that did work, your workflow and processes will need time and iterations to adapt to this new aspect of business. So start small with AI, see how everything reacts and then plan for the next phase. Phases should be small and objectives/benefits split across phases.

3. Make it Objective:

List out the metrics and RoI. Be brutal here. All AI effort should address tangible and quantifiable business goal. What are the success criteria? What is the measurable impact? Who is responsible for what? All these will make analysis post the execution very objective and make next steps clear. Also you would have a yard stick to check things every now and then.

4. You Have Enough Data:

If it’s your first tryst, I am sure this is thrown a lot at you “We do not have enough data”. It’s the proverbial half a glass of water.

It might be true that you actually have less than sufficient. But, unfortunately, if that is so, you will never have any more of it. So either AI/ML has to work on the data you have or it doesn’t apply to that use case for you. Or grab data. Solution architects have to devise a neat and seamless way of collecting data. You may start with a nearly rule based system and then grab data along the way. In any case, data grabbing strategy has to be a part of the AI solution.

And finally –

5. You can Understand AI:

It is not out of the world. It’s just maths and engineering. You can always understand how it works. So give it a shot understanding as a maths problems solved using programming and engineering. And don’t be overwhelmed by the jargon. Rather, hear and read it so many times that you get used to it. Eg. Stochastic means random. Probability means chance. Build a glossary of terms and have an online dictionary handy. It’s workable.

As concluding remarks, I would say – We don’t want to stay behind in this fast moving market, nor we want to jump the gun. Start small but get started. Also, be ready to encounter a few unforeseen challenges along the way. But hey, you have overcome many such in the past.

I hope the above points help alleviate concerns and apprehensions around AI integrations and show a way out. Feel free to comment with your questions and share this with other business stakeholders if it seems relevant.

Leave a Reply