Business strategy in the AI era: what the data says, and what 67 industrial homepages reveal
Business strategy is the set of choices that decide how a company will win: which customers it serves, what it offers them, and what it deliberately will not do. Michael Porter gave the modern version of this idea in a 1996 Harvard Business Review essay, where he defined strategy as "the creation of a unique and valuable position, involving a different set of activities" and separated it from operational effectiveness, which "means performing similar activities better than rivals perform them."[1] Generative AI has made that distinction more practical than it has been in decades.
This essay brings together four bodies of evidence. The first is survey data on how fast AI tools spread among workers and companies. The second is a field experiment on what those tools do to the quality and speed of professional work. The third is a count of scholarly research on generative AI, which shows where attention has gone. The fourth is our own scan of 67 US mid-cap manufacturing and supply-chain homepages, which measures how clearly industrial companies state what they sell and to whom. We close with a framework, the Choice Ladder, and a method for writing a strategy on one page.
Idea in brief
The problem. AI tools reached most companies at once. 88% of respondents in McKinsey's 2025 global survey said their organization regularly uses AI in at least one business function, but only 39% reported any profit impact at the enterprise level.[3]
Why it happens. A tool that helps everyone equally cannot be the reason one company wins. In a field experiment, GPT-4 lifted below-average consultants by 43% and above-average ones by 17%, narrowing the gap between them.[4] Porter called the result competitive convergence: "races down identical paths that no one can win."[1]
The gap (our data). Industrial companies often do not state their choices. In our scan, 60% of homepage title tags named a product, service or customer group, but only 16% of homepage headlines did. Half the headlines were slogans.
The answer. Put the choices first and the tools second. Write a diagnosis, name who you serve, list what you will not do, check that your actions fit, and only then decide where AI goes. AI scales a choice. It cannot make one.
Key findings
The essay in 3 minutes- Finding 128%
The tools spread fast.
The share of employed US adults who had used ChatGPT for tasks at work rose from 8% in March 2023 to 28% in early 2025.[5]
- Finding 26%
Use is common, value is rare.
88% of McKinsey's 2025 respondents reported regular AI use, 39% reported any EBIT impact at the enterprise level, and about 6% qualified as AI high performers.[3]
- Finding 380%
Efficiency is the default goal.
80% said their companies set efficiency as an objective of their AI initiatives. The companies seeing the most value often also set growth or innovation as objectives.[3]
- Finding 460%
AI levels execution inside its frontier and hurts outside it.
In the BCG experiment, AI raised task completion from 82% to 91% to 93%. On a task chosen to sit outside the frontier, correct answers fell from 84.5% to 60% and 70%.[4]
- Finding 50.7%
Research follows productivity, not advantage.
Scholarly works mentioning generative AI rose from 147 in 2022 to 27,218 in 2025. Of the 2025 works, 15.4% also mentioned productivity or efficiency, and 0.7% mentioned competitive advantage.
- Finding 616%
Industrial headlines rarely state a position.
Of 67 readable homepages, 11 (16%) had a headline that named what the company sells or who buys it. 34 (51%) had a slogan.
Chapter 1 · 1 min
What business strategy is
Strategy is a set of choices that fit together and that a rival would struggle to copy. Porter's essay gives the core test: "A company can outperform rivals only if it can establish a difference that it can preserve."[1] Richard Rumelt, writing in McKinsey Quarterly in 2011, gave the structure. He called it the kernel of good strategy: "A diagnosis: an explanation of the nature of the challenge," "A guiding policy: an overall approach chosen to cope with or overcome the obstacles identified in the diagnosis," and "Coherent actions: steps that are coordinated with one another to support the accomplishment of the guiding policy."[2]
Terms used in this essay
| Term | Meaning |
|---|---|
| Business strategy | The set of choices about where a company competes and how it wins, including what it will not do. |
| Operational effectiveness | Doing similar activities better than rivals: faster, cheaper, with fewer defects. Porter's term.[1] |
| Trade-off | A choice where doing more of one thing means doing less of another. Porter: "The essence of strategy is choosing what not to do."[1] |
| Competitive convergence | Rivals copying each other's improvements until their strategies look alike.[1] |
| Kernel | Rumelt's three parts of good strategy: diagnosis, guiding policy and coherent actions.[2] |
| Jagged frontier | The uneven boundary between tasks AI does well and tasks it does badly, even when they look equally hard.[4] |
Rumelt lists four hallmarks of bad strategy: "the failure to face the challenge, mistaking goals for strategy, bad strategic objectives, and fluff." He defines fluff as "a restatement of the obvious, combined with a generous sprinkling of buzzwords that masquerade as expertise."[2] Keep that definition in mind for our scan below.
Chapter 2 · 1 min
The evidence, part 1: everyone got the same tools at once
Pew Research Center has asked the same question of US adults since March 2023: have you ever used ChatGPT for tasks at work? The answers trace how quickly the tool entered ordinary jobs.
Figure 1. ChatGPT went to work fast
Percent of employed US adults
| Survey dates | Employed adults who have used ChatGPT for work |
|---|---|
| March 2023 | 8 |
| July 2023 | 12 |
| February 2024 | 20 |
| February to March 2025 | 28 |
Share of employed US adults who say they have used ChatGPT for tasks at work, in percent. Source: Pew Research Center, American Trends Panel, topline published June 2025.
In the same survey, 34% of all US adults said they had ever used ChatGPT, up from 18% in July 2023 and 23% in February 2024.[5] One chatbot is only a slice of AI use, but the tools are clearly becoming standard equipment, not a private edge.
Company surveys show the same spread, and a gap behind it. McKinsey's 2025 global survey drew 1,993 respondents in 105 countries between 25 June and 29 July 2025.[3]
Figure 2. Nearly everyone uses AI. Few see the profit
Percent of respondents, 0 to 100
Share of survey respondents, in percent. AI high performers are respondents reporting EBIT impact of 5% or more from AI and "significant" value; McKinsey puts them at "about 6 percent of respondents". Source: McKinsey, The state of AI in 2025, 5 November 2025.
Use rose from 78% a year earlier to 88%.[3] Nearly two-thirds of respondents said their organizations had not yet begun scaling AI across the enterprise. The report's key findings also note that the companies seeing the most value "often set growth or innovation as additional objectives" beyond efficiency.[3] That is Porter's distinction showing up in survey data. Efficiency is operational effectiveness. Growth and innovation require choices.
Tip: List the three things your business does better than rivals. For each, ask whether a competitor could match it within six months using off-the-shelf AI tools. If yes, it is an operational edge, not a strategic one.
Chapter 3 · 2 min
The evidence, part 2: AI levels execution, unevenly
The strongest evidence on what AI does to professional work comes from a pre-registered field experiment by researchers from Harvard Business School, MIT, Wharton and Warwick with Boston Consulting Group.[4] 758 consultants, about 7% of BCG's individual contributors, were randomly assigned to work with no AI, with GPT-4, or with GPT-4 plus a prompt engineering overview.
Figure 3. Inside the frontier, AI helped everyone finish more
Percent of tasks completed, 0 to 100
Average share of 18 consulting tasks completed, for tasks inside the frontier of GPT-4's abilities, in percent. Source: Dell'Acqua et al., Harvard Business School Working Paper 24-013, 2023.
On those tasks, consultants with AI completed 12.2% more tasks on average, worked 25.1% faster and produced results rated more than 40% higher in quality. The gains were largest for the weakest performers: 43% for those below the average threshold, against 17% for those above it.[4] AI narrowed the distance between good and average execution.
The researchers also designed one task to sit just outside what GPT-4 could do well.
Figure 4. Outside the frontier, AI made answers worse
Percent correct, 0 to 100
The paper reports the two AI conditions at 60% and 70%, an average fall of 19 percentage points.
Share of consultants who reached the correct recommendation on a task chosen to sit outside GPT-4's frontier, in percent. The paper reports the two AI conditions at 60% and 70%, an average fall of 19 percentage points. Source: Dell'Acqua et al., 2023.
The paper also reports that ideas from the AI groups were better on average but showed "a marked reduction in the variability," which "might lead to more homogenized outputs."[4]
For a strategist, the lesson is that AI raises the floor of execution, pulls people toward similar answers, and can mislead capable people on problems that look routine but are not. Deciding which problems are which is a strategic judgment.
Limits of the evidence
- The BCG tasks were chosen to fit, or deliberately not fit, a 2023 model. Newer tools have a different frontier.
- McKinsey's figures are self-reported by survey respondents, not measured from company accounts.
- Pew measures ever having used one product, not depth or frequency of use.
If you want an outside read on whether your current offer shows real choices or just faster execution, the Brave New audit looks at how buyers read your positioning and where it blurs.
Chapter 4 · 1 min
The evidence, part 3: where the research attention went
To see where attention has gone, we counted scholarly works on generative AI in OpenAlex, a free, open index of research papers, books and reports, on 3 November 2026.
Figure 5. Research on generative AI exploded after 2022
Number of works, 0 to 30,000
Number of works in OpenAlex whose title or abstract contains the phrase "generative AI", by publication year. Source: Brave New Media count of the OpenAlex works API, 3 November 2026.
We then checked how many of the 2025 works also mentioned a set of strategy and productivity terms.
Figure 6. Productivity gets the attention. Advantage barely registers
Percent of 2025 generative AI works, 0 to 20
"Strategy" often has a technical meaning in these works.
Share of the 27,218 works published in 2025 whose title or abstract contains "generative AI" that also contain each term, in percent. Source: Brave New Media count of the OpenAlex works API, 3 November 2026.
The word "strategy" is common, but in technical papers it often means a training or prompting strategy rather than a business one. "Competitive advantage", a phrase with only one meaning, appeared in 188 of 27,218 works. Read with care, this suggests that the research conversation, like McKinsey's respondents, is mostly about doing things faster. Very little of it asks how a company becomes harder to copy.
Chapter 5 · 2 min
Original research: do industrial companies state their choices?
A strategy that buyers never hear does little for growth. The clearest public trace of a company's position is its homepage: the title tag, written mainly for search engines, and the main headline (H1), written for the people who arrive. So we checked whether US mid-cap industrial companies use those lines to say what they sell and to whom.
The sample is the 91 companies in the S&P MidCap 400 and S&P SmallCap 600 classed under eleven manufacturing and freight sub-industries, the same sample used in our earlier AI-readiness scan. On 3 November 2026 we fetched each homepage once with a script that named itself. 67 returned a readable page. We coded each title tag and each first headline into one of four categories, using a written rule (see Methodology). The full coding sheet is published with this essay so anyone can check or dispute a call.
Figure 7. Title tags name the offer. Headlines mostly do not
Percent of 67 homepages, 0 to 100
Title tag: names offer or customer 59.7% (40), slogan 11.9% (8), not a positioning line 23.9% (16), missing 4.5% (3).
Homepage headline: names offer or customer 16.4% (11), slogan 50.7% (34), not a positioning line 16.4% (11), missing 16.4% (11).
Share of 67 readable homepages of US mid-cap industrial companies, by category, in percent. "Not a positioning line" covers a company name, navigation text, a news item or a recruiting line. "Missing" means no title or H1 in the raw HTML. Source: Brave New Media scan, 3 November 2026.
40 of 67 title tags (60%) named a product, service or customer group. Search engines reward that, so most companies do it. The headline a human reads first did so in 11 of 67 cases (16%). 34 headlines (51%) were slogans that named no product, service, industry or customer group.
The slogans are revealing when lined up. "Engineering a Stronger Tomorrow." "Engineering Tomorrow's Solutions." "Powering a Clean Future." "Lightweighting to Propel the Future." "Impossible Ends Here." "Expect the Best." Each belongs to a real company with real products, and none of them would tell a buyer in a new market what those products are.
The 11 headlines that did name the offer show the alternative. "Optimized for 48V power delivery." "Stronger shearwall. Simpler installation." "Digital Manufacturing for Early Prototyping to Final Production." "One logistics partner for all your freight." A buyer knows at once whether they are in the right place.
This does not measure whether these companies have a good strategy; many have clear positions their sales teams state every day. It measures whether the most-read line on the website carries that position to a stranger. Rumelt quotes a statement agreed by a computer maker's executives, "committed to providing high-quality products and services and being a leader in data processing," and calls it "not a strategy" but "a political outcome."[2] A headline that could belong to any company in the sample is the public version of that problem.
Chapter 6 · 1 min
A framework: the Choice Ladder
The evidence points to an order of work. Each rung depends on the one below it, and AI belongs at the top, where it multiplies choices already made, rather than at the bottom, where it replaces them.
| Rung | Question it answers | Evidence | What to check |
|---|---|---|---|
| 5. Scale | Where does AI multiply the choice? | AI lifted task completion from 82% to 93% inside its frontier.[4] Value-seeing companies set growth goals, not only efficiency.[3] | AI on necessary but undifferentiated work, and on carrying one clear message to more markets |
| 4. Fit | Do the actions support each other? | Rumelt's coherent actions; Porter's activity fit.[1][2] | Three or four moves, each making the others stronger |
| 3. Trade-offs | What will we not do? | "The essence of strategy is choosing what not to do."[1] | A written "we will not" list where each item costs something |
| 2. Position | Who do we serve, with what? | Only 16% of industrial homepage headlines in our scan named an offer or customer. | One line a stranger could repeat |
| 1. Diagnosis | What is the real obstacle? | Rumelt: bad strategy fails to face the challenge.[2] | Two sentences naming the main problem, not a goal |
Most companies start at the top: they buy tools and look for uses. The data suggests why that disappoints. 88% use AI, but 39% see any enterprise-level profit impact.[3] Starting at the bottom gives the tools something to scale.
Chapter 7 · 2 min
How to write a one-page business strategy in six steps
Step 1: Write the diagnosis in two sentences
Describe what is happening and the main thing blocking you. "Sales are flat" is a symptom. "We win new clients easily, but most leave within a year because we look like every other supplier" is a diagnosis.
Step 2: Name who you serve, narrowly
Pick the customer you are best for, and describe them so tightly that a stranger could find them: industry, size, situation and the problem they have now. The Hive's guide to market segmentation is a good refresher.
Step 3: Write your "we will not" list
This is the step that matters most. Write three to five things you will not do: customers you will not take, services you will not offer, places you will not sell, prices you will not match. Each item should hurt a little. If it costs nothing to give up, it is not a trade-off.
Step 4: Choose your guiding policy
In one sentence, state the approach you will take to beat the obstacle from Step 1. For example: "We will become the specialist for X, and compete on depth of knowledge, not on price." This is where your positioning comes from. The guiding policy is the business decision. The positioning is how customers hear it, and your homepage headline is where many of them hear it first.
Step 5: Decide where AI goes, and where it does not
Now bring AI in, and only now. Put it on work that is necessary but not distinctive: first drafts, scheduling, reporting, summaries, routine replies. Keep people on the work your strategy says makes you different, and on problems that may sit outside the frontier, where the BCG consultants who leaned on AI got worse answers.[4] The Hive's explainer on what AI marketing is covers the marketing side of this split.
This is also where AI earns its place as a growth tool. Once the choices are made, an AI flow can take the same clear message to more channels, more markets and more buyers without adding headcount at the same rate. That is the kind of flow we build at Brave New Media.
Step 6: List three or four coherent actions
Write the few moves you will make next quarter. Each should support the others and the guiding policy. If an action does not follow from the diagnosis, cut it, even if it is a good idea.
Tip: Read the finished page aloud. If it takes more than two minutes, or a new employee could not tell you what you would turn down, keep cutting.
Which step to do first
| Step | Choice Ladder rung | Effort | Likely impact | When |
|---|---|---|---|---|
| Two-sentence diagnosis | Diagnosis | Low | High | This week |
| "We will not" list | Trade-offs | Low | High | This week |
| Rewrite the homepage headline to name offer and customer | Position | Low | Medium to high | This month |
| Map where AI goes and where it does not | Scale | Medium | Medium | This quarter |
| Three or four coherent actions | Fit | Medium | High | Each quarter |
Chapter 8 · 1 min
A real case: one airline chose, one tried to do both
Porter's main example is Southwest Airlines in the 1990s. "Southwest does not offer meals, assigned seats, interline baggage checking, or premium classes of service." Those choices let it turn planes around at the gate in 15 minutes.[1] Continental Airlines tried to match it on some routes with Continental Lite while staying full-service everywhere else, keeping travel agents, a mixed fleet, baggage checking and seat assignments. Porter's account: "Late flights and cancellations generated a thousand complaints a day," and "The airline lost hundreds of millions of dollars, and the CEO lost his job."[1] Adopting the same AI tools as every rival, while changing nothing about what you choose, is a quieter version of the same mistake.
Chapter 9 · 1 min
A worked example of a one-page strategy
This case is made up, but it's typical. Say Marta runs a 40-person machine shop that makes precision metal parts. She quotes on everything: automotive brackets, farm equipment, furniture fittings, one-off repairs. Her rivals use the same quoting software and the same AI tools to write quotes, so quotes look alike and price decides most jobs.
| Part | Marta's answer |
|---|---|
| Diagnosis | We quote for every kind of buyer, so we win jobs at thin margins and lose them to anyone a little cheaper. |
| Who we serve | Makers of laboratory and medical equipment, with tight tolerances and strict paperwork for every batch. |
| We will not | Bid on high-volume automotive parts. Take one-off jobs below a minimum order. Match a quote that skips our inspection steps. |
| Guiding policy | Be the supplier equipment makers trust with inspected, traceable parts, and compete on reliability, not price. |
| Where AI goes | Quote drafts, scheduling, formatting inspection reports, follow-up emails. Not the engineering review or the choice of which buyers to take. |
| Coherent actions | Train the team on the documentation these buyers ask for. Keep a traceability record for every batch. Rewrite the homepage headline and quotes for equipment makers only. Use an AI flow to reach equipment makers in two neighbouring countries. |
The pieces fit. The narrow customer makes the training worth the cost, and the training makes the traceability record credible, which gives her a reason not to match cheaper quotes. Because the message is narrow, it travels to the same kind of buyer in a new market. A generalist rival could copy one piece, but copying all of it would mean dropping its automotive work.
Chapter 10 · 1 min
What this means for leaders
For a manufacturing or supply-chain business, AI changes the cost of execution far more than the cost of choosing. Quoting, documentation, reporting and outreach all get cheaper, for you and for every rival. The experiment and the survey data agree that those gains are real and widely shared, which is exactly why they will not hold a margin on their own.[3][4]
What AI does change is the reach of a good choice. A company that knows who it serves and what it will not do can now carry that position into more markets, more languages and more channels than its headcount would once allow. It starts with a line as small as the homepage headline, which in our scan said what the company sells in only one case in six.
If you want a clear outside view of how your choices come across to buyers, and then an AI flow that takes them to more markets, start with the Brave New audit. You can also put a specific question to the Brave New Q&A.
Chapter 11 · 1 min
Common mistakes
- Calling an AI rollout a strategy. "We'll use AI to be more efficient" is an operational goal, and 80% of companies in McKinsey's survey share it.[3]
- Mistaking goals for strategy. "Double revenue" says where you want to land, not how, or what you will give up.[2]
- Straddling. Copying a rival's best move while keeping all your old commitments, like Continental Lite.[1]
- Hiding the position. A strategy the homepage headline does not state is invisible to every buyer who has not met your sales team.
- Defining yourself by your product. Theodore Levitt's warning about marketing myopia applies: describe the need you meet, or a new tool will make your product look ordinary.
Once the choices are clear, a sharp unique selling proposition turns the guiding policy into one line a buyer can repeat, and distinctive brand assets help people remember you when they are ready to buy.
Chapter 12 · 1 min
Methodology
Research counts. On 3 November 2026 we queried the OpenAlex works API (api.openalex.org), a free index of scholarly works, with a title-and-abstract search for the phrase "generative AI", grouped by publication year from 2019 to 2025. We repeated the query with each added term: strategy; productivity OR efficiency; differentiation OR positioning; and the phrase "competitive advantage". OpenAlex keeps adding records, so counts for recent years will rise over time. Matching is on words, not meaning.
Homepage sample. All 91 companies listed on Wikipedia's S&P MidCap 400 and S&P SmallCap 600 constituent pages under the GICS Industrials sector in eleven manufacturing and freight sub-industries. Official websites came from Wikidata (property P856), corrected by hand for five companies.
Fetch. On 3 November 2026 a script requested each homepage once with the user agent "BraveNewResearch/1.0 (+https://brave-new.com; research)". It read raw HTML only, ran no JavaScript and did not retry blocked requests. 67 of 91 returned HTTP 200. 17 returned 403, 3 returned 202 and 4 returned another status; these 24 are excluded.
Coding. For each of the 67 readable homepages we coded the title tag and the first non-empty H1 headline into one of four categories. Names offer or customer: a stranger reading only this line could say what the company sells (a product or service category) or who buys it (a named industry or customer group). Slogan: an aspiration, value or claim with none of those. Not a positioning line: a company name, "Home", navigation text, a news item, a recruiting line or a single-product promotion. Missing: no title or H1 in the raw HTML. One analyst coded all lines with this rule; there was no second coder, so borderline calls are possible, and every call is published for checking.
Limits. Homepage lines are a narrow window on strategy. Pages built by JavaScript may show a different headline to a person than to our script. The sample is US-listed mid-cap industrials only. We did not measure sales, margins or what buyers think.
Chapter 13 · 1 min
Frequently asked questions
What is business strategy in simple terms?
Business strategy is the set of choices that decide how a company will win: which customers it serves, what it offers them, and what it deliberately will not do. A good strategy names the main obstacle, sets an approach to beat it, and lists a few actions that support each other. Goals and budgets alone are not a strategy.
Does AI change how companies should think about strategy?
Yes, mostly by raising the stakes. In a field experiment, GPT-4 helped weaker consultants most, narrowing the gap between good and average work. McKinsey's 2025 survey found 88% of organizations use AI but 39% report any enterprise profit impact. When execution gets cheap for everyone, the choices about who you serve and what you refuse matter more.
What is the difference between strategy and operational effectiveness?
Operational effectiveness means performing similar activities better than rivals: faster, cheaper or with fewer errors. Strategy means choosing a different set of activities to deliver a unique mix of value. Michael Porter argued in 1996 that best practices spread quickly, so operational gains do not last. AI is a fast-spreading best practice.
What should a one-page strategy include?
A short diagnosis of the main challenge, a narrow description of who you serve, a list of things you will not do, one guiding policy, a note on where AI and people each fit, and three or four actions for the next quarter. If you cannot read it aloud in about two minutes, it is too long.
How should a manufacturer use AI in its business strategy?
Decide the position first, then use AI to scale it. Put AI on quoting, documentation, reporting and outreach, and keep people on engineering judgment and the choice of customers. Then state the position where buyers look first: in our scan of 67 industrial homepages, only 16% of headlines said what the company sells or who buys it.
Your next step
This week, take one sheet of paper and write only Step 3: three things your business will no longer do. Then open your homepage and read the main headline as a stranger would. Does it say what you sell and to whom? If not, rewrite it from your "we will not" list, because the things you refuse are often the clearest description of what you are.
If you want a team to test how that position reads to buyers, and an AI flow that carries it into new markets, start with the Brave New audit.
Brave New Media is a marketing agency that builds AI flows to help companies scale their ideas, grow their customer base and expand into new markets, with experience across supply chain and manufacturing.
See also
- Positioning in marketing, Brave New Hive
- Market segmentation, Brave New Hive
- Unique selling proposition, Brave New Hive
- Marketing myopia, Brave New Hive
- The Brave New Business Journal, weekly thinking on strategy, risk and money in real markets
References
1. Porter, M. E. "What Is Strategy?" Harvard Business Review, November to December 1996. hbr.org/1996/11/what-is-strategy
2. Rumelt, R. "The perils of bad strategy." McKinsey Quarterly, 1 June 2011. mckinsey.com
3. McKinsey and Company (Singla, A., Sukharevsky, A., Hall, B., Yee, L., Chui, M. and Balakrishnan, T.). "The state of AI in 2025: Agents, innovation, and transformation." 5 November 2025. mckinsey.com
4. Dell'Acqua, F., McFowland, E., Mollick, E., Lifshitz-Assaf, H., Kellogg, K. C., Rajendran, S., Krayer, L., Candelon, F. and Lakhani, K. R. "Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality." Harvard Business School Working Paper 24-013, 2023. mitsloan.mit.edu
5. Sidoti, O. and McClain, C. "34% of U.S. adults have used ChatGPT, about double the share in 2023." Pew Research Center, 25 June 2025, with topline for American Trends Panel Wave 164. pewresearch.org
6. OpenAlex works API, queried 3 November 2026. openalex.org
7. Brave New Media. Homepage positioning scan of 91 US mid-cap industrial company websites, 3 November 2026. Data and coding sheet published with this essay.