Why Feasibility Studies Are Still the Gold Standard – For Any Endeavor – Even AI
Let’s start with a reminder of the definition. A feasibility study is a researched document that provides an objective assessment of whether a project, business idea, or system development is likely to succeed before major resources are committed.
Of all the issues I have had while working in this field, first and foremost is that “objective” part. No, you can’t change the price of windows if the market in the relevant area says they will cost X. You cannot decide to skip the implementation consult on a complex project because “we’ll figure it out.” Hint – it is not cheaper. Send your people to school, or hire help.
Structuring a useable study requires specific non-negotiable elements.
- Define the goal to reach or the problem to solve.
- Define the steps required.
- Do the research to get real numbers on all the components.
- Define a timeline and determine the impact on the numbers.
- Assess the probability of success and search for savings that do not impact adversely on that success.
Step one: define the goal or the problem.
This may change. Perhaps you start out thinking you are going to build a building with six floors, but number crunching tells you the costs for adding, say the top three floors, adds carrying costs to build-out and structural changes to support the desired design. Maybe a three story comes online faster, at a lower costs, and cash flows at a higher return.
Define the problem. Don’t jump into the newest thing without understanding what problem it will solve. Rather than “we want AI!” first answer the question “for what?” The only way to build a successful implementation is to nail down what the goal is. Then the numbers flow from there.
Step two: define a baseline of steps required to get there.
What type of conceptional design should be the first look. What are the permitting costs and how long does it take? Is the land under option, or accruing interest and taxes while the process goes through? What materials should be used and what are some sources? Who is the contractor and what is their schedule? Every step takes time and costs money. Build that into the model or the project will come in over budget and may never cash flow against the investment.
Let’s talk AI. See step one. You cannot make a determination of what solution will meet your needs if you have no notion of what those needs are. Talk to staff, talk to management, talk to colleagues, and THEN seek out an AI consultant that will sell you what you need and not just an expensive entertainment system for your staff (or a disaster waiting to happen at light speed). Put the time in you need here or I promise you will pay for it later. Once AI is interwoven into your process and you are unhappy with the results, it is not a flip of a switch to start over. Always consider what data is being shared. If working with people’s personal finances, health care, or anything considered PII, do you really want that information exposed to what is basically public tool? Find someone who understands AI security.
Step three: do the research.
Here is where AI can shine, as long as you follow a few rules. I remember the days when gathering data on utility rates, taxes, local rents, city ordinances, and a multitude of required components for successfully costing a project was painfully tedious. Do it right here – do it honestly and objectively – or pay for it later. Perhaps in failure. And always, always check the sources.
We are told that AI can replace staff, however, what is the cost of implementation? Does anyone know what the staff does including whatever task is being “replaced.” What is the cost of subscriptions, upkeep, and licensing. See step two – talk to staff. Maybe an interesting goal is to reduce overtime – not just start with layoffs.
Then there is the ghost in every budget meeting – opportunity costs. If funds are invested in this project today, what programs that currently contribute to cash flow and profit margins will lose funding and what might that impact be? Time to market is crucial, regardless of what the pursuit is. If you don’t have the cash to meet the market at the right time and place, it no longer matters how fancy your AI process is or how inspiring your building is. What will you miss if you invest now, or is there a better target date?
Understand the commitment. Anytime a complex implementation is set in motion it is absolutely non-negotiable to hire the consultants that know which switches need to be set and why before you start. Or, have your staff trained to do it right. Pay the price here, or suffer years of broken functions and lost market share.
Case in point: Group Health decided to change their system for mail order prescriptions about a decade ago. This was not a small system. When it went live it went toes up. No prescriptions were going out, no one could figure out where they were, and they were caught in a spiral of not being able to process new prescriptions until the old could be canceled. This was apocalyptic and a massive health risk to their patients. When customer service was finally contacted (seems their lines were rather busy) the only option was to go to a physical, Group Health Pharmacy, and work it out. Envision the line. Customer service, already overwhelmed, could only reply – it worked in the trial. At the very least they should have staffed up customer service before they flipped the switch. Don’t be that company. [Note: I haven’t met a software engineer yet that believed that a system would function perfectly on implementation day and they planned for it. Maybe I’m just picky who I work with.]
Step four: define the timeline.
Use all that data that has been collected and determine a realistic time line. Remember – time is money. Don’t assume your team can be faster than anyone on the planet. Someone is, but they might not have hit the bumps you do. They might not be working with the same legacy systems. Calculate the costs. All of them. Interest to carry working capital, subscriptions, staff, transition, retraining, or reassigning. Ikea took on AI, then retrained the existing staff to meet new demands. Make it visual. Set milestones. Determine if a different start date will free up current staff to add to the success of the project. [Hink: there are times of year when accountants are dead-eye focused on deadlines. If you want a smooth transition, work within that structure and not against it).
Step five: reassess.
Step back and get the big picture. Are there points in your plan that can be moved to reduce carrying costs. Are there alternatives that are better suited to the project, now that you have a complete plan in place and not just a vague notion of where you are going? Do you need all of the aspects of your chosen AI coming out of the gate, or is it better to stage implementation making sure the foundation is solid? Test the numbers, the resources, and outcome. Is this the best solution to get to the end goal?
This is the hardest step because sometimes, the best decision is to walk away. Save the ideas for a later date when you might have more information, or carve out the parts that work for now and can be expanded later. If the job has been done correctly, you can trust what you see and make a solid business decision. One that creates value, not a never-ending fix it project that sucks the life out of the success you have built.
