Services· AI strategy·Where does my budget go furthest?
Vanguard · Guide
Which of your AI initiatives deserve to continue, which to scale and which to switch off?
We have AI pilots in five departments and budget to scale two.
Your initiatives ranked by return, by risk and by what it costs to keep them running, with the list of the ones worth switching off and the reason for each.
AI Prioritization (by Function or Sector)Send this page to whoever decides
This sounds like you if
- Every department brings its pilot and all of them look reasonable one at a time.
- There is approved budget and more initiatives than money to scale them.
- The diagnostics you get describe companies that look nothing like yours.
Use this today, without hiring anyone
Four questions that drop half your list in an afternoon. Run them on every initiative before you ask anyone for a proposal:
- How much volume is there really? Count the times per month. A process that runs twenty times does not pay for an automation, however painful it is.
- Is the exception the rule? If every case is resolved differently, there is no process to automate: there is judgment, and judgment does not automate cheaply.
- Who keeps it running on day two hundred? Not who builds it. If you cannot write down a name, the initiative has already failed.
- Will this process still exist as it is in eighteen months? Automating something that is about to change is the most common way to spend a budget well on the wrong thing.
And a fifth question that changes depending on where you look at the portfolio from:
- If you run operations: multiply monthly volume, timed unit duration and rework rate. Annualized and at loaded hourly cost, that is the number that opens the budget.
- If you answer for people functions: rank by consequence for the person before return. Whatever affects someone's access to a job moves up the risk list even if it saves little.
- If you run sales: pull the deals lost after stage two and put on them the hours they consumed. That is the cost of qualifying badly.
- If you are in banking or finance: ask of every initiative whether you could reconstruct one specific decision for a supervisor in under an hour.
- If you are in health care or insurance: check first what happens if the system gets someone's access to a service or a coverage wrong.
The ordering that changes the result: rank by where the value chain loses money, not by which department is most enthusiastic or which initiative is technically easiest.
How we solve it
The method, not the promise.
- Your value chain gets mapped down to the level where time, money or quality is lost. That level is deeper than what is usually documented.
- The initiatives get inventoried, including pilots in flight and what some departments are already doing without permission.
- The lens of your function or your sector gets applied, which is where the order changes: consequence for the person, the stage where deals fall, or reconstruction for the supervisor.
- Impact gets estimated with a range and an open assumption, and who keeps each initiative running on day two hundred gets assessed.
- The ranking gets delivered with written, reproducible criteria, with each initiative marked to continue, scale or switch off.
What you receive
- The map of your value chain down to the point where the money is lost.
- The ranking with its written criteria, reusable without us.
- The top three initiatives described at the level at which they can be contracted.
- The list of what is worth switching off, with its reason.
- The reading specific to your function or sector, with whatever specialist is needed when the matter is regulated.
The proof that applies here
- Portfolio prioritization under budget constraints across eleven countries, with the drop list in writing.
- Six Sigma Black Belt.
Before you hire
This work ranks and drops. When the matter is labor, privacy or regulation specific to your sector, the interpretation comes from a specialist brought in before the work starts.
Ordering a portfolio is a judgment call, and a judgment call is signed by whoever makes it. What is delivered here is reproducible, so you can run it again the following year without calling anyone. That is the proof that what you bought was judgment.
Of your AI initiatives in flight, how many have a name attached who will answer for them six months from now?
In the cycle