Smart, Practical AI
The promise of Artificial Intelligence to transform work is real and already delivering measurable results. The teams we have led have applied AI to reduce human effort in some processes by up to 40%, freeing up people to higher-value work.
And these results do not require large, expensive IT projects – they are within the reach of leaders and teams with access to approved desktop AI platforms like Microsoft 365 Copilot or Chat GPT.
But how? This Harvard Business Review article offers a sensible strategy to start to capture those results:
AI can be implemented in a focused / productive way with a strong return on investment (start with the problems to be solved), OR haphazardly and scattered (start with the solution), needlessly consuming excess time and effort.
Common Pitfall: Start with the Solution (AI)
Develop Solutions, then look for for problems to solve
- Bloated, over-scoped projects: more time, effort and cost
- Solutions may not solve underlying problems
- Slower results - relief slow to arrive, if it arrives at all
- Lost momentum - enthusiasm wanes as projects drag on, “tax” on slow work
- Teams disengaged and sceptical
Who Benefits?
- Consulting firms who want you to spend more money with them, who have profit motives in over-scoping your work.
Optimal Strategy: Start with the Process and its Business Problem
Define your business problems and causes, then develop solutions
- Less time, effort, and cost to solve
- More realistic solutions that provide measurable results
- Faster results - faster relief; momentum - avoid the "tax" on slow work
- Free up time to develop internal capacity to improve further
- Engaged teams
Who Benefits?
- You and your teams – actual measurable relief faster
- Your clients, citizens – faster, easier processes that deliver better
- Your CIO – because you’re not over-developing AI solutions, capacity remaining to serve more IT clients.
"A problem well-stated is a problem half-solved" - Charles Kettering
If you first understand the business problems that you’re trying to solve, and their root causes, then you are better positioned to determine what kind of solution is best: an AI solution, or a process improvement solution.
Case Example: AI or Process Improvement Solution?
In several processes where we have helped clients use AI and process improvement together, taking a systems-thinking approach to understanding problems and their root causes led to performance breakthroughs.
For example, consider a Procurement process where users submit forms with missing, incorrect and unclear information, causing back and forth, wasting effort and ruining the service experience.
Solution 1: Chatbot
A popular solution is to implement a chatbot to help users fill out application forms correctly and completely.
However, if you assessed why the user struggles to complete the form correctly, you would learn that the current form:
- Is written in “service provider-speak”, not the language of the typical user, using unfamiliar terms and jargon;
- Is wordy, with a layout that isn’t intuitive or “at-a-glance” so users lose patience when they don’t understand what the form is about or how it works;
- Collects information that is not required by the process.
Solution 2: Fix the Form
- Conduct focus groups with typical users to highlight issues and provide suggestions for redesigning the form to ensure that 95% of first-time users fill it out correctly on their first attempt, without assistance.
- At the same time, identify and delete the form’s fields that are not used in the process, simplifying and lightening the ask of the client.
The latter solution directly addresses the cause of the back and forth. At the same time, this solution does not exclude AI’s considerable value.
Using AI’s analysis of emails to and from users, you can measure and summarize the most frequent client errors, omissions and clarifications, and use this data to redesign and error-proof the form. Further, you can use AI to measure if the first-time-through performance has improved once the new form is in place.
Overview of the Approach
By following variations of the method below, our clients have achieved measurable improvements within weeks, satisfied with the efficiency and effectiveness of their improvement project.
From real-life Procurement transformations, where the total process elapsed time was reduced from 3-5 months to just 3-5 weeks, our data suggests:
AI Solutions
Process Improvement Solutions
Effort Time Savings
40%
60%
Elapsed Tiime Savings
30%
70%
AI Solutions
Effort Time Savings = 40%
Elapsed Time Savings = 30%
Process Solution Example
Effort Time Savings = 60%
Elapsed Time Savings = 70%
Some examples of AI and Process Improvement solutions.
AI Solution Example
- State the potential AI solution as a “user story” to then create prompts to test the solution using the AI platform:
Problem: Busy Procurement Officer spends a lot of time sifting through documentation and assessing which procurement method is appropriate.
- Design the detailed AI solution
- Test, reflect, and adjust
- Communicate, train, implement more widely, and build into standard work
- Measure and quantify the benefits
Process Solution Example
- Write the process change solution experiment as a hypothesis:
Problem: There are 8-10 back and forth emails and drafts between the client and the Procurement Officer to refine and finalize the RFP documents, taking an average of 8 weeks and 20 hours of effort.
- Design the detailed process solution
- Test, retrospective, adjust
- Communicate, train, implement more widely, and build into standard work
- Measure and quantify the benefits
When to NOT use AI
- Avoid relying on AI 100% when there are ethical, legal or financial implications: for instance, a qualified human should always review an RFP before posting it publicly.
- AI makes internal tasks easier and faster but presents the risk of creating bottlenecks in the end-to-end process, creating delays for the client. The optimal solution is to combine the time-savings of AI with the coordination and collaboration of process improvement, so that both employees and clients benefit.
- As a solution to a symptom, not a root cause – unless the AI solution is lightweight and temporary to buy time to solve the root cause of a complex problem.
In this content area, we cover:
- Understand basic desktop AI tools (e.g. MS 365 Copilot) and key opportunities to start using them to create process flow and liberate human effort.
- Develop effective problem statements to outline and rank the core business problems to be addressed.
- Use advanced root-cause analysis focus solutions on causes of business problems, instead of only addressing symptoms.
- How to avoid improvements that simply do “the wrong thing, faster.”
- Identify opportunities where AI and Process Improvement can together create greater combined impact than either approach on its own.
- Understand where AI and Process Improvement each create the greatest value, and where not to apply each approach.
- Craft effective prompts for AI experimentation using common desktop AI tools (e.g. 365 Copilot), and user stories for advanced AI experiments requiring programmed/coded solutions.
- Crafting effective process improvement hypotheses (A3’s).
- Apply AI techniques to enhance Lean/Process Improvement tools and methods for streamlining processes.
- Use an AI-enhanced DMAIC method to facilitate improvement projects and achieve breakthrough measurable results.
- Use the Plan-Do-Check-Adjust cycle to conduct experiments to quickly learn and improve both process- and AI-based solutions.
- The human side of implementing AI and Process Improvement.
Find the AI Approach That Fits the Problem You’re Trying to Solve (Westerman, Ramsbotham, and Farronato) Harvard Business Review February 6, 2024.
Copyright © 2022 Lean Agility Inc. All rights reserved. Powered by QuickSilk.