1. Setting realistic business goals
We have identified problems that make a good fit for AI and how businesses need to calibrate their expectations from an AI project.
Such an informed understanding plays a vital role in project scoping, which is the focus of this video.
2. Goals setting
AI initiatives come to life starting from setting the right goals. Building an AI chatbot can be a high-priority initiative to improve customer service.
3. Broad goals are not effective!
One goal would be to “improve customer service”, but its broad nature doesn't easily translate into a technical challenge.
4. SMART framework
To refine such goals, frameworks like SMART, which stands for specific,
5. SMART framework
measurable,
6. SMART framework
achievable,
7. SMART framework
relevant,
8. SMART framework
and time-bound metrics to set objectives.
9. SMART framework
Reframed using this approach, the objective could become something like – "Build an AI chatbot with an 80% accuracy by next quarter to cut the response time by 10% leading to improve customer service."
10. Specific and measurable
It carries the specific motivation, such as to improve customer satisfaction and experience by reducing the time to respond to their queries.
Practical goals are measurable. For instance, achieving an 80% model accuracy that results in a 10% faster response time compared to a non-AI system provides clear metrics for success.
11. Relevant and time-bound
It is relevant as “building bot” is aligned with the business goal of improving customer experience through automated response.
As every project has allocated time and resources, the timeline for building a chatbot over the next quarter provides a clear timeframe for planning, execution, and evaluation.
12. Achievable
Lastly, it must be achievable. But what does achievability mean in the first place?
It is a two-step process – the first being feasibility workshops where the AI strategy leans on different teams' expertise to assess the solution's viability.
The second one is benchmarking, which involves either a comparison with the existing approach to managing the operations
13. Achievable
or building a prototype or smaller version of the project, generally called PoC, which will be explained in detail in the next chapter.
14. PoC
In other words, it's recommended to first roll out the initial results of any AI project, the chatbot in this case, to a smaller audience to gauge its performance
15. PoC
and make adjustments
16. PoC
before full-scale deployment.
17. The fixed timeline issue
One common concern among AI strategists is that AI projects are often exploratory and require extensive research.
In such cases, enforcing a goal achievable within a fixed timeline sometimes becomes ambitious.
18. The fixed timeline issue
This issue is magnified when solutions require extensive experimentation,
new tech stack integration,
deployment,
monitoring systems, and
synergies among stakeholders for updating and maintaining these complex systems.
19. The stakeholders
The overarching goal of an AI project is a sum total of multiple sub-goals from all stakeholders involved, including business,
20. The stakeholders
data,
21. The stakeholders
model development,
22. The stakeholders
and the MLOps team.
23. The stakeholders
An AI strategist aligns the availability and goals of the participating teams to ensure the last mile delivery.
24. The kill switch
AI strategists face some complex decisions on when to terminate a project. This decision is challenging given the sunk cost of the time and resources invested in the project.
25. The ROI indicator
Hence, deciding these conditions upfront where the return on investment, or ROI, offers clarity is essential.
Suppose the expected returns don't compensate for the project delays and cost overruns, mainly if these delays reflect a lack of confidence in successful project execution.
In that case, ROI is a good indicator to make an informed decision alongside project sponsors.
26. Let's practice!
To summarize, setting realistic business goals is crucial to managing expectations while also charting out an exit plan.
To further cement this understanding, let's help some organizations set their AI goals.