“I would not hand the board a percentage until I can name the exact workflow the dollar comes from. Most leaders I see can feel AI should make money, but only about one in four can point at the work that produces it. Audit what you already own first, then fund one 90-day pilot tied to revenue per employee, with kill criteria at day 60.”

“I wouldn't set a budget line until you separate what must change your cost structure from what merely sounds like AI. At OpenAI, the leaders who got value fastest weren't the ones with the biggest pilots; they were the ones who could say exactly which expensive process intelligence would replace.”

“I would not hand the board a single AI budget number until the leadership team agrees which workflows actually change and who owns the result. At Cisco I watched a $1.1B innovation portfolio drift when every pilot had a sponsor but no stop rule. The money that matters now is the part you are willing to lose on three fast tests, not the headline percentage.”

“I would skip the big budget line item and give one curious person real latitude to walk across departments and fix manual workflows that are already costing you money. You will learn more in a month than in six months of board strategy decks, and then you will know what is actually worth funding.”

“I would not set the budget number until the board can name the three use cases that own a metric. At AWS and EQUS I watched the same trap: companies fund AI like infrastructure, then discover 70% of agent initiatives stall at adoption and only 26% can see the token meter in real time. Budget for experiments with a kill switch and a business outcome, not a percentage of IT spend.”

“I would not set the budget as a percentage of spend until you have a ranked list of use cases with a measurable outcome attached to each one. At Apple, I consolidated more than 70 AIML use cases across Media Product Operations before leadership would align on funding, and the ones that survived were those with a clear before-and-after metric, not the ones that sounded transformative.”

“If you are a traditional B2B software company adding AI features, my shorthand is 5 to 8% of budget, maybe up to 20% if it is genuinely transformative. If the AI is so good the product sells itself, you can spend 60%.”

“I’ve shipped 24 AI systems and I still think the board is asking the wrong question. Budget percentage is a vanity metric until you know which three workflows actually move revenue or cut cost. I would carve out enough to run three small, bounded pilots tied to real P&L lines, not one big strategy deck.”

“I would not hand the board a budget percentage until I knew where the constraint is. I just watched that with a friend's business: AI surfaced plenty, but the real bottleneck was a process AI could not touch, so spending more would have been theater. Map one use case that changes EBIT, fund that, then decide if the rest is worth it.”