5 questions to ask before you start an AI project
We don’t need to convince many people that data is important. We don’t even need to try very hard to convince clients they should take a data-driven approach to their strategies, implement hypothesis-based problem solving, and test/iterate their way to better outcomes.
What’s harder is figuring out where to start and how to get quick wins under the belt while wading through a sea of buzzwords and “5 steps to… ” guides that really don’t get you any closer to getting started, leave you more confused, and somehow, on an email list that floods your inbox with subpar content three times a week.
That said, here are 5 steps questions you should ask before starting your next data/AI project. And since all of your projects should be data driven, these are the 5 questions you should ask before you start any project.
1 - What are you trying to achieve?
It sounds obvious, I know. But, maybe because it seems so obvious that a lot of people skip this step.
Stepping back and simply stating “What are we actually trying to do?” is a powerful first step to get the team on the same page.
You can break that down further: “What problem do you have that you are trying to solve?” and “What is the goal and what tests should we deploy first?” These questions also align the team on what data should be collected to help validate (or de-validate) your hypothesis.
Without knowing what you want to achieve you might succeed at getting the right answer to the wrong question. We talk about it in an episode of our podcast Sound BITES.
I overuse this phrase but here it is anyway: If you don’t know where you are going, any direction will get you there.
2 - What are you doing now?
Conduct a quick audit of what you’re currently doing and how.
You need to determine if:
you currently have the data to analyze or if you need to go looking for it
there is another stakeholder who you need to work with to get the right outcome
this a greenfield project where you need to start collecting fresh data to get moving on the path
Understanding what activity is currently taking place, what data has been and is being collected, understanding where it lives and how to get it can save you a ton of duplication of effort.
Many times the answer to this question is: “Nothing yet” and that’s okay, too. You gotta start somewhere.
3 - What does success look like?
Knowing up front what a win looks like and establishing how success is being measured can make or break your project.
For some projects, the success criteria may be as simple as “establish a baseline to measure future activity.”
In Agile teams this strategy makes sense because you are constantly trying new things and seeing if it was successful relative to the previous test. Whatever your objective, you should align the team around minimum success criteria so everyone knows what sport we are playing and what a goal looks like. Doing so also prevents success drift; not every project will be a home run, especially as you are trying new things and learning, but retroactively moving the goal posts can destroy your credibility.
Test, learn and iterate, but don’t lie to yourself about the outcome or else that iterate phase won’t be nearly as impactful.
4 - What are you going to do with the results?
Step 1 - collect data
Step 2 - ???
Step 3 - profit
What action will you take based on what the data is telling you?
If the answer is “I don’t know” then you need to go back to the drawing board and consider if you need to collect that piece of data at all. This is probably the most overlooked item on the list. We work so hard to align the team on the objective and success criteria to start collecting data, that we somehow forget to define what action we should take.
Before you collect anything you should state:
If the data says X, we will try this next
If the data says Y, we will do this next
This ultimately removes emotion later when you invariably get a result you didn’t expect.
Too many times we’ve seen great projects that require a ton of resources to get off the ground, stall when the team doesn’t know what action to take based on the data and instead keeps doing the same thing while watching a dashboard that has become meaningless while they hope the direction of the chart changes at some point.
Hope is an awful strategy and an even worse tactic.
5 - What resources and internal buy-in do you have or need?
Speaking of those resources, what resources will your project require?
What stakeholder support do you need from inside and outside of your organization? Are you in a large organization with strict data governance and the data team buried in the IT branch of the org? It might be time to get to know that team for more than just fixing your misbehaving computer (did you turn it off and back on?)
Internal buy- in is much easier with the right relationships, and it can make or break any initiative. Asking the right questions so you can get a sense for what’s possible is a good place to start, and helping the cross-functional teams understand what you are trying to accomplish (see #1) as well as what you are going to do with the results (see #4) generally helps with buy-in.
Beyond internal resources you very likely have external vendors, strategy teams, data analytics as a service providers, and agencies in your solution stack to get on the same page
Bonus
We recommend you start with smaller projects that could have potential for quick success with low risk.
Are there any quick wins we can get to demonstrate value, get momentum, get a feel for the process, and scale into stickier problems with higher risk? You might find that your project may not require any AI at all to get early value and quick wins while giving you a path to layer in some of that robot magic later.
“Quick wins” are defined as high impact, low effort projects. Identifying a few of these will buy a lot of confidence with the team and stakeholders, and should give some freedom to tackle more complicated projects later.
TLDR (too long, didn’t read)
To boil it down to a tweet: Have a clear objective, know what success looks like, understand that everything is iterative and quick wins will lead to long term success.
What are you trying to achieve?
What are you doing now?
What does success look like?
What are you going to do with the results?
What resources and internal buy-in do you have or need?
While some of this may seem obvious, more often than not one of these 5 questions go unanswered at the start of high-stakes data projects leading to poor results. But that won’t be you. No way. You are already 80% more likely to succeed by asking and answering just a few questions before you start your journey.
Have any best practices for your AI and data projects? Let us know we would love to hear what you think and have you on our podcast to talk through it.