Tomas Halloran leads a nine-person inside-sales team at a building-materials distributor. For a year he had been letting his reps try AI tools on their own: one used a chatbot to draft follow-up emails, another pasted call notes into a summarizer, a third built quote spreadsheets with an AI assistant. It looked like progress, but when his director asked "so what has AI actually done for your numbers this quarter?" Tomas had no answer. The tools were scattered, the wins were anecdotal, and nobody could say whether any of it moved the needle. That Friday he sat down and admitted the real problem: his team had AI activity but no AI strategy. This lesson is the process he used to fix that, and it is one you can run for your own team in an afternoon.

What This Lesson Covers

An AI strategy is a short, deliberate set of choices about which problems you will use AI to solve, in what order, and how you will know it worked. It is not enterprise transformation and it is not the CEO's job. At the manager level it is far more grounded: you are deciding how your team or department will get real value from AI over the next two or three quarters, and what you are deliberately choosing not to chase.

You will learn the difference between a strategy and a wish, how to find the two or three use cases worth your time, how to sequence adoption so the team is not overwhelmed, how to set goals using OKRs (a simple goal-setting method explained below), how to measure whether AI is actually paying off, and how to carry your team through the predictable resistance that comes with any new way of working. We follow Tomas the whole way.

Strategy Is Choosing, Not Doing More

The trap Tomas fell into is the most common one for managers: mistaking activity for strategy. Buying tools, running pilots, and encouraging people to "experiment with AI" feels strategic. It is not. A strategy is defined by what it rules out as much as by what it pursues.

A useful definition for a manager: a strategy is a coherent set of choices about which problems you will solve, what capability you will build, and where you will put your limited time and budget to create a real advantage for your team. Notice the word choices. If your plan is to do everything, you have no strategy; you have a backlog. The discipline is deciding that customer-quote turnaround matters this quarter and that, say, AI-assisted competitor research does not, even though it would be interesting.

A strategy is not "we are going to use more AI." A strategy is "we are going to cut quote turnaround from two days to two hours, and we are not touching anything else until that works."

Start with a problem, never with the tool. The weakest team AI plans begin with "we want to use AI" and then go hunting for somewhere to apply it. The strong ones begin with "this specific thing hurts" and then ask whether AI can help. Tomas's team had three real pains: quotes took two days to turn around, follow-up emails slipped through the cracks, and call notes were never written up properly. Those pains, not the tools, became the spine of his strategy.

Four Things People Call Strategy That Are Not

Before Tomas could write one, he had to get clear on what he was actually producing, because four other artifacts routinely get mislabelled as strategy and each of them left him with a different kind of gap.

• A plan is not a strategy. A plan is the execution vehicle for a strategy. Tomas had a rollout schedule long before he had a reason for it.
• A goal is not a strategy. A goal is an outcome you want. A strategy is how you intend to get there. "Faster quotes" was a goal; nothing about it told him what to build or what to skip.
• A list of initiatives is not a strategy. Initiatives are the building blocks. Three reps each trying a different tool was a list, and a list makes no choices.
• An aspirational vision is not a strategy. Vision is where you want to be. Strategy is the route. "We will be the most responsive distributor in the region" is a fine ambition and useless as a decision rule.

What a real strategy does is answer five questions, and if you cannot answer them you are still holding one of the four artifacts above. What competitive advantage are we trying to achieve? What capabilities do we need to build to achieve it? Where should we focus our resources given our constraints? What are we explicitly choosing not to do? And how will we know if the strategy is working? Tomas wrote those five questions at the top of a blank page and refused to leave any of them empty.

Your Altitude: Domain Strategy, Not Enterprise Strategy

It helps to see the difference in altitude between what a chief executive writes and what you write. At the top of a company, an AI strategy sounds like this: we are investing in AI to automate routine tasks across the organization, freeing people for higher-value work and reducing our cost structure by a set percentage. Broad, directional, and correctly so, because it has to cover every function.

A manager's version is narrower and far more specific. A customer service leader might write: in our service function we are investing in AI assistants to handle the majority of routine inquiries, cutting response time from two hours to fifteen minutes and freeing the team for the complex issues that need human judgment, which improves satisfaction and lowers cost per inquiry. That sentence names the domain, the change, the numbers, and the benefit.

Tomas's version was shorter still, because his domain is nine people and their quotes. That is the point. Your AI strategy is scoped to the function you lead, the operational or competitive advantage you are trying to create there, the capabilities you need to build, the investment you are prepared to make, and the measures you will judge it by. Anything wider than that belongs to somebody above you, and anything narrower is a task list.

The Elements of a Strong AI Strategy

A strategy that survives contact with a real quarter has six recognisable parts. Tomas worked through them in order, and you can treat them as a checklist for whatever you draft.

A clear problem or opportunity statement

Every strong strategy opens with a business problem, not with a tool. Notice how much decision-making pressure a well-written problem statement carries: "customer response time is a competitive disadvantage, customers wait more than two hours while competitors reply in fifteen to thirty minutes." Or "our content process is a bottleneck: we can produce ten pieces a month and we need twenty-five, and the current process cannot scale." Or "the quality of our customer communication is inconsistent, some people write clearly and some do not, and it shows in how customers see us." Each of those orients every choice that follows. Tomas's was blunt: quote turnaround of two days was costing deals to faster competitors.

An honest read of your own context

What works for one organization fails in another, so a good strategy accounts for where you actually stand rather than copying a case study. Tomas walked through six context questions. What are the dynamics of your industry, and does AI genuinely create an advantage there? What do your customers expect, and do they care about the specific improvement you are considering? How mature is your organization with AI already, early or advanced? What does the technology landscape offer, and what is realistically viable for you? What talent and capability do you have to implement and manage these systems? And what does your regulatory environment allow or restrict? In a distribution business with price-sensitive buyers, Tomas concluded that speed was noticed and that nothing in his industry stopped him from drafting quotes with AI as long as a human signed them off.

A named competitive advantage

Competitive advantage is simply a distinguishing capability that lets you outperform others in a specific dimension, and advantages from AI tend to fall into five families. A cost advantage lowers your cost of operations so you can compete on price or margin. A speed advantage lets you move faster than competitors, to market or to the customer. A quality advantage makes your output better or more consistent. A customer experience advantage builds loyalty and differentiation. An innovation advantage gives you capabilities or offerings that others simply do not have.

You cannot chase all five at once, so choose. Tomas chose speed, with quality as a constraint he refused to trade away. That single choice told him which use cases mattered and which were merely interesting.

A capability roadmap

A capability is an organizational competency you need in order to execute, and a roadmap is simply the sequenced plan for building those competencies over time. Capabilities are usually about people and process rather than about the tool, which is why buying software so often fails to deliver the strategy attached to it.

Consider a strategy built on AI-assisted customer service. To make it real you need knowledge management so the assistant is grounded in your own domain, data systems to track how it performs and where it improves, team expertise to manage and keep improving it, integration so it connects to the systems you already run, and quality assurance so the customer experience stays consistent. Tomas mapped the same five onto quoting: the current price list had to be somewhere the tool could use, he needed a simple record of turnaround times, one rep had to own and refine the prompt, the draft had to reach the quoting system without retyping, and every quote needed a review step before it left the building.

Part of the roadmap is deciding what you build yourself and what you rely on a vendor to provide. Be deliberate about that line rather than discovering it halfway through.

Explicit investment and prioritization

Budget, talent, and time are all finite, so a strategy has to say where they go. The test of seriousness is whether the document names what you are not funding this period. We will invest in these three things; we will not invest in those three, at least not now. Tomas put a modest training budget and two hours a week of his own time behind quoting and call notes, and wrote "not this quarter" next to competitor research. Those exclusions are what made the rest achievable.

Metrics that match the advantage you chose

Your measures should trace directly back to the advantage you are pursuing, otherwise you will end up reporting numbers that prove nothing. If your advantage is cost, measure cost per unit of output and how it compares to alternatives. If it is speed, measure time to deliver and how that compares. If it is quality, measure defect rates and satisfaction. If it is customer experience, measure satisfaction, retention, and recommendation. Tomas had chosen speed, so turnaround time became his headline number, with an accuracy measure guarding it.

Step One: Take an Honest Inventory

With the elements clear, the work becomes a sequence. Tomas started where every strategy should start, with an unflattering look at the present. Before deciding where to go, write down where you already are: which AI tools are in use anywhere in your function, what business problem each of them is supposedly solving, what measurable impact they have had, what barriers or problems have surfaced, and what gaps remain.

His inventory took forty minutes and was sobering. Three tools in use, none of them agreed on, one paid for on a personal card, no impact anyone could quantify, and an obvious gap around quotes, which was the thing customers actually complained about. That inventory did more to justify the strategy to his director than any slide he could have built.

Step Two: Look in the Five Places Opportunities Hide

Next, ask where AI could help, and look systematically rather than waiting for inspiration. Five lenses will surface almost everything worth considering. What do customers complain about, and could AI address it? What operational bottlenecks limit your capacity, and could AI raise that ceiling? What quality problems keep recurring, and could AI make output more consistent? What drives your cost structure, and could AI reduce it? And what is simply slow, where AI could accelerate the work?

For each candidate that emerges, size it before you fall in love with it. How big is the opportunity, in customers affected or money involved? How likely is AI to actually solve it? What barriers stand in the way, technical, organizational, data-related, or skills-related? And what investment would it require? Tomas ran his three pains through those four questions and discovered that the one his reps were most excited about, competitor research, was the smallest and least tractable of the set.

Finding Your Two or Three Use Cases

Resist the urge to tackle ten things. A team can absorb two or three new ways of working at once; beyond that, nothing sticks. To find the right two or three, Tomas listed every candidate task and scored each on four plain questions:

• How much does it hurt today? How much time, money, or customer frustration does this problem cause across the team each week?
• How well-suited is it to AI? AI is strong at drafting, summarizing, reformatting, and first-pass analysis. It is weak at judgment calls and anything needing private context it cannot see.
• How hard is it to adopt? Does it need new tools, new data access, or sign-off from IT and legal, or can the team start Monday with a tool they already have?
• Does a win build momentum? An early, visible success makes the next change easier. A slow, invisible one drains belief.

There is a fifth question worth adding once you have more than a couple of candidates: does this advance the advantage you named? Strategic alignment is what separates a genuinely useful improvement from a merely convenient one. A use case can save real hours and still belong on the later list because it does nothing for the advantage you chose to pursue.

His shortlist came out clearly. Quote drafting hurt the most, suited AI well, and a faster quote was something customers would notice immediately, so it was the obvious first move. Call-note summarization was easy and low-risk, a good confidence-builder. Follow-up email drafting was useful but lower-stakes, so it became the third. Everything else, including the competitor-research idea one rep loved, went on a "later" list, written down so it felt parked rather than rejected.

Sequencing: One Win Before the Next

Order matters as much as selection. Doing all three at once would have meant three half-learned habits and no clear signal about what worked. Tomas sequenced deliberately. He led with the lowest-risk, fastest-feedback use case to build the team's confidence and his own evidence, then moved up to the higher-value, higher-effort one.

He started with call-note summarization, even though quotes were the bigger prize. The reason: it was nearly impossible to get wrong, the team felt the benefit in a day, and it taught everyone the basic rhythm of working with an AI assistant before the stakes got higher. Two weeks later, with the team comfortable, he rolled out quote drafting, the use case that actually moved revenue. Follow-up emails came last, once the first two were habit. Sequencing this way meant each new ask landed on a team that already trusted the approach, instead of a team being asked to swallow everything cold.

Writing the Strategy Down

For your top priority, put the strategy on a single page and answer five things: what advantage you are creating, what capabilities you need, what you will invest, how you will measure success, and on what timeline. Vagueness here is the most reliable predictor of a strategy that never happens, because nobody can act on a sentence that could mean anything.

A poor strategic statement reads: "we are going to be more efficient with AI." Nobody can argue with it and nobody can execute it. A better one reads like this: we will implement AI-assisted customer service to cut response time from two hours to thirty minutes and free the team for complex issues; we will judge success on customer satisfaction, on response time, and on cost per inquiry, each against a named target; and we will invest a defined budget in tools and training over six months. Every clause is checkable.

Tomas wrote his in about twenty minutes once the inventory and the opportunity sizing were done. The point of the page is not elegance. It is that six weeks later, when someone asks why a rep is spending Tuesday morning rewriting a prompt, the answer is already written down.

Setting Goals With OKRs

Activity without a target drifts. To keep the strategy honest, Tomas set goals using OKRs, which stands for Objectives and Key Results. The method is simple. The Objective is a short, plain statement of what you want to achieve and why it matters; it is meant to be a little ambitious. The Key Results are two to four specific, measurable outcomes that tell you whether you hit the objective. The objective is the destination in words; the key results are the numbers that prove you arrived.

A good key result is a number with a target and a deadline, not a task. "Use the AI quote tool" is a task. "Cut average quote turnaround from 48 hours to 4 hours by end of Q3" is a key result, because you can measure it and you cannot fool yourself about it.

A Worked Example: Tomas's First AI OKR Set

Here is the actual OKR set Tomas wrote for his team's first two AI use cases, covering one quarter. Note that the key results mix the value he wants (faster quotes, more follow-through) with the adoption he needs to get there (the team actually using the tools), because a value target with zero adoption is just a hope.

Objective: Make our team noticeably faster and more reliable for customers by putting AI to work on quotes and call follow-up.
KR1: Cut average quote turnaround from 48 hours to 4 hours.
KR2: Write up and log call notes within the same day for 90% of sales calls, up from roughly 40% today.
KR3: Reach 8 of 9 reps using the AI quote-draft step on at least 80% of their quotes by the end of the quarter.
KR4: Hold or improve quote accuracy, measured as the share of quotes needing no manual correction, at 95% or better.

Look at how KR4 guards against a bad outcome. Speed is worthless if accuracy collapses, so Tomas added a quality guardrail: he would only count the strategy a success if quotes got faster and stayed correct. This is a habit worth keeping. Whenever an AI use case optimizes for speed or volume, pair it with a key result that protects quality, or you will win the metric and lose the customer.

At the end of the quarter the team landed KR1 at about 6 hours (good progress, not the full 4), hit KR2 at 88%, hit KR3 at 7 of 9, and held KR4 at 96%. None were perfect, and that is the point of OKRs: they are scored, not pass-fail, and a 70-to-90% landing on an ambitious set is a genuine win. Tomas now had a real answer for his director, backed by numbers.

Measuring Whether It Actually Paid Off

Adoption numbers tell you people are using the tool. They do not tell you it was worth it. To measure real value, tie each use case back to one of four kinds of benefit and capture a simple before-and-after.

• Time saved: hours per week the team gets back. Tomas estimated call write-ups dropped from about 15 minutes each to 4, saving the team roughly 6 hours a week he could redirect to actual selling.
• Speed: how much faster the customer gets a response. Quote turnaround going from 48 to 6 hours was the number customers actually felt.
• Quality or consistency: fewer errors, more uniform output. Logged call notes meant any rep could pick up a colleague's account without a cold start.
• Revenue or retention: the hardest to attribute cleanly, but the one leadership cares about most. Faster, more reliable quoting plausibly nudged win rates; Tomas reported it as a contributing factor, not a sole cause, because honest attribution builds more credibility than a confident overclaim.

The cheap discipline that makes all of this real is capturing a baseline before you start. Tomas spent one week logging current quote times and note-completion rates before he rolled anything out. Without that baseline he would have had no honest way to claim improvement later.

Communicating It and Building Alignment

A strategy is only useful if the people who have to execute it understand it and agree it is worth doing. Tomas booked thirty minutes with the whole team and walked through seven things, in this order. What business problem are we solving? Why do we think AI is the right answer to it? What are we trying to achieve? How will we measure whether it worked? What is required from each of you? What timeline are we on? And what support or resources do you need to make it happen?

That last question is what turned the session from an announcement into a conversation. Two reps raised concerns he had not considered, one about customers spotting a template-sounding quote and one about who owned the price list. He adjusted the rollout on the spot for the first and assigned an owner for the second. Build alignment through exactly this kind of exchange: listen to the concerns, adjust where the concern is right, and be clear that you remain committed to the direction even where you cannot resolve everything. Alignment is not unanimity. It is a team that knows the plan and knows why.

Managing the Change: The Change Curve

A strategy on paper is not a strategy in practice until people change how they work, and people resist that even when the change helps them. A useful map here is the change curve, a simple model of the emotional stages people move through when a new way of working arrives: first denial ("this won't really change anything"), then resistance ("this is extra work, the old way was fine, what if it makes mistakes?"), then exploration ("okay, let me try it on a couple of quotes"), and finally commitment ("I would not go back"). The manager's job is to expect this curve and help people across it, not to be surprised that day-one enthusiasm is low.

Tomas hit resistance fast. His most experienced rep, Gwen, was openly skeptical: she was sure the AI quotes would be wrong and embarrass her in front of customers. Arguing would have entrenched her. Instead Tomas did three things that map onto moving someone from resistance to exploration. He named the fear honestly ("you are worried it will get a price wrong and you will look bad, that is fair"). He shrank the risk by making the AI draft a starting point she always reviewed, never an auto-send, so she stayed in control. And he let a small win do the convincing: he asked her to use it on just three low-stakes quotes that week. By the third she had caught one small AI error in review (which proved the review step worked) and saved twenty minutes on the other two. Gwen moved herself from resistance to exploration faster than any mandate could have pushed her.

Two more change-curve habits served Tomas well. He found and fed his early adopters, the reps already excited, by letting them demo their wins in the team meeting so the enthusiasm came from peers rather than from the boss. And he was patient with the timeline, because commitment is earned over weeks of small successes, not declared in a kickoff email.

Three Tradeoffs You Will Have to Decide

Strategy is choosing, and three choices come up for almost every manager. There is no universally right answer to any of them, only an answer that fits your context and constraints.

Build versus buy

Do you build a custom AI solution tailored to exactly how you work, or buy something off the shelf? Building gives you a better fit and more control at the price of a larger investment and a longer timeline. Buying gets you deployed faster and cheaper, with less control and an imperfect fit. Most managers land in the same place: buy standard tools for standard problems, and consider building only where the problem is genuinely novel. Tomas bought, because quote drafting is not a novel problem.

Speed versus perfection

Do you deploy something good enough now and iterate, or refine it until it is right before anyone touches it? Speed gets you to value sooner and lets real use teach you what to fix. Perfection lowers the risk of a visible failure and raises launch quality, at the cost of a much longer wait for any benefit at all. Most managers choose speed: ship a working version, measure it, improve it. Tomas's first quote prompt was mediocre and his fourth was good, and he only learned the difference by using it.

Centralized versus decentralized

Does one group own AI adoption across the function, or does each team adopt what it likes? Centralizing gives you a consistent approach, shared learning, and real governance, at some cost in flexibility. Decentralizing gives you fast innovation and team-specific fits, at the cost of consistency and with a genuine risk of fragmentation. Most managers settle in the middle: the core tools are centralized and agreed, and teams are free to innovate around the edges. That middle ground is exactly what Tomas built when he standardized on one assistant for quotes while leaving the "later" list open to whoever wanted to experiment.

Three Strategy Traps to Avoid

Copying another team's playbook. Hearing that another department automated its reporting with AI and deciding to do the same ignores that their problems, data, and skills differ from yours. A manager who reads that a competitor is using AI for a particular application and immediately commits to the same thing usually discovers that the customer base, the capabilities, and the constraints were all different, and the fit is poor. Diagnose your own pains first; borrow ideas, not conclusions.

"AI everywhere" with no problem behind it. Declaring that the team will use AI for everything spreads effort thin and produces a pile of half-used tools. There is no clear problem driving it, no specific advantage being pursued, and no realistic view of where AI actually adds value, so the results come back scattered and mixed. Anchor every use case to a specific pain and a specific target, or drop it.

Strategy that never reaches execution. A beautifully reasoned one-page strategy that sits in a doc while nothing changes is the most common failure of all. Months of thinking produce a thoughtful document, then the team does not prioritize it, resources never get allocated, and nothing about the work changes. The strategy is only real once it is communicated, resourced, measured, and revisited. Tomas reviewed his OKRs in a fifteen-minute slot every two weeks; that small ritual is what kept the plan alive instead of filed away.

Executing, Measuring, and Adjusting

Execution is the unglamorous half. You work the plan, you watch the measures, and you change course as evidence arrives. A strategy is not fixed: as you learn, you adjust it. But there is a difference between adjusting and abandoning, and the difference is usually just difficulty. Do not throw out a strategy every time a week goes badly. Stay committed through the hard middle, and change direction only when the evidence, not the discomfort, tells you to.

This is where managers actually shape the future of their organizations. The ones who win with AI are not the ones with the best tools. They are the ones with the clearest strategies: they know which problems they are solving, they know which advantage they are creating, they allocate resources on purpose, they measure relentlessly, and they adjust based on what they learn. It is not flashy work and it has nothing to do with chasing the newest trend. It is strategic clarity and disciplined execution, and it is what produces an advantage that lasts.

Put This Into Practice

Four exercises will take this from reading to doing. Start by identifying the top three business problems in your own domain that AI might help solve, and for each one assess how big the opportunity is, how likely AI is to solve it, what investment it would take, and what advantage a solution would create.

Then pick one of those problems and write the one-page strategy for it: the business problem, the competitive advantage, the capabilities you need, the key investments, and your success metrics. Keep it to a page on purpose.

Third, name a strategic tradeoff your organization is genuinely facing, whether that is build versus buy, speed versus perfection, or centralized versus decentralized, and document the tradeoffs on each side before deciding which option fits your context.

Finally, go and interview a peer manager about their AI strategy. What is their strategic priority, how did they choose it, how are they measuring success, and what have they learned that they did not expect? Half an hour with someone one department away is often worth more than a week of reading.

Reflection

Sit with three questions before you write anything. What is the clearest competitive advantage AI could create in your domain, and what would that advantage actually let you do? If you could build only one AI-related capability in your organization, which would it be, and why that one? And what is the single biggest constraint limiting your ability to execute an AI strategy right now, whether budget, talent, or organizational readiness, and how would you go about addressing it?

Closing Thought

Strategy is the bridge between aspiration and execution. It is where you decide what to pursue and what to ignore, where you commit resources, and where you begin to shape what your part of the organization becomes. Build yours on clear problems, connect it to a real advantage, make your investment choices explicit, measure without flinching, and adjust as you learn. Tomas started that Friday with nothing but scattered activity and a question he could not answer. A quarter later he had a strategy, numbers to back it, and a team that understood why.

Key Takeaways

• Strategy is choosing, not doing more. A team AI strategy is defined by what you deliberately exclude as much as by what you pursue; "use more AI" is not a strategy.
• Start from a problem, never a tool. Name the specific pain that hurts your team each week, then ask whether AI can help, not the reverse.
• Pick two or three use cases and sequence them. Lead with a low-risk, fast-feedback win to build confidence, then move up to the higher-value, higher-effort one.
• Set OKRs with a quality guardrail. Pair value and adoption key results, and add a key result that protects quality so you do not win speed while losing accuracy.
• Capture a baseline before you start. One week of measuring the current state is what lets you make an honest before-and-after claim later.
• Plan for the change curve. Expect denial and resistance; name the fear, shrink the risk, and let a small real win convert skeptics far better than a mandate.
• Keep the strategy alive. A short, regular review of your OKRs is what turns a one-page plan into actual change instead of a filed document.
• Answer the five strategy questions. Advantage, capabilities, resource focus, what you are not doing, and how you will know it worked. If any is blank, you have a plan, a goal, or a vision, not a strategy.
• Match your metrics to your chosen advantage. Cost, speed, quality, and customer experience each demand different measures; reporting the wrong ones proves nothing.
• Decide the three tradeoffs on purpose. Build or buy, speed or perfection, centralized or decentralized. Choose for your context and constraints, not because of what others are doing.

Frequently Asked Questions

Isn't AI strategy the CEO's job, not mine? Enterprise-wide AI strategy is leadership's domain, but the strategy for how your team gets value from AI is squarely yours. You know your team's real pains, your customers, and your constraints better than anyone above you. A focused team-level strategy, picking two or three use cases and measuring them, is exactly the altitude a manager should own.

How many use cases should I start with? Two or three, not ten. A team can build two or three new habits at once; past that, nothing sticks and you cannot tell what is working. Park the rest on a written "later" list so they feel deferred rather than rejected, and pull from it once your first wins are solid.

What if I cannot cleanly prove AI caused a revenue change? Then do not claim it did. Report the metrics you can attribute confidently, time saved, faster turnaround, fewer errors, and describe revenue or retention as a likely contributing factor rather than a proven cause. Honest, modest attribution builds far more credibility with your leadership than a confident overclaim that someone can poke a hole in.

My team is resistant. Do I just mandate it? Mandates produce compliance, not commitment, and compliance quietly reverts the moment you stop watching. Use the change curve instead: name people's fears, reduce the risk by keeping a human review step, and let early adopters and small visible wins do the persuading. Genuine adoption you build this way survives without your enforcement.