K.I.S.S. Please...
The majority of inbound consulting requests we get come from potential clients who have several good ideas around AI implementation within their organization, but are not equipped to begin to map out a path forward. They are overwhelmed with knowing where to start, how to get from idea to a real deliverable in this space.
Keep it simple... stu... I mean stakeholder.
Identifying the potential use cases may seem easy, but getting collective buy in on where to start is typically challenging. My advice would be to have an ideation session on potential objectives of an AI Proof of Concept (PoC) Bring together a good mix of Executive sponsorship, operational stakeholders, frontline workers, design engineers and a strong facilitator.
The task is to list out objectives to potentially achieve. At this step you are identifying a task or routine function that can be automated or enhanced with AI. Naturally each objective should be mapped to an over arching Business Goal, something core to the company. Then clarify the exact metric by which success can be measured.
Keep the descriptions to as few of words as possible, reduce the complexity to spur on ideation. We are so often paralyzed by seeming complexity, it blocks our collaboration and hinders results.
Now that you and your team have decided on the objectives to pursue in your AI PoC. This is another opportunity to keep it simple. If your objective is to reduce cost, identify the baseline and your target improvement level. If your objective is to reduce processing time, Identify the current time suck for the task and state your target optimal duration. If your goal is to reduce error rates, define the current error rate, and what your future goal will be.
This activity gets all participants on the same page about the expected outcomes of the AI Implementation. Speak now or forever hold your peace! If you feel the targets are out of line or if the baselines to not reflect true reality now is the time to make it known.
The next simple step is sharing timelines, this typically will come from a project leader or engineering service delivery manager. Input from all parties, and commitment to timeframes and participation availability are paramount in mapping out the timeline.
Recommended by LinkedIn
One thing is clear, if you do not innovate quickly, iterate efficiently, fail fast, and adapt - your AI project will get lost in the wash. The foundation systems and models evolve so quickly, you must be mindful of how you integrate to be "upgradeable" easily.
The rough flow of any AI implementation project is as follows:
I advise pushing this process along at a 1 to 2 week clip per phase. If you cannot execute that quickly something, is out of balance. If the project is too large, timeliness of results will suffer. If the project is decently sized but under staffed, it may take too long and suffer from engineer burnout. Think about sizing your PoCs to this level of leanness to give yourself better chances of success. This approach also opens up the lane to iterations of multiple PoCs by keeping labor loads more manageable.
Remember to keep it simple. The more direct and clear the approach angle, the more likely you are to land the plane. When your team begins the process of developing a Proof of Concept for your AI implementation remember these tactics.
Whiteboard wisely,
Benjamin Justice