Posted by
Sam Beddall
AI can make you astonishingly productive and still not improve the business. Most marketers have already felt some version of this. You use AI to draft a blog post, summarise a meeting, build a quick prototype or generate 20 campaign ideas before lunch, and the output appears so quickly that it feels like progress.
Then comes the awkward question. Did any of it make the marketing budget go further, or did it simply generate more activity?
AI is not short of impressive demonstrations. The tools can now build apps, write code, generate copy, create slides and analyse data in a fraction of the time those tasks used to take. But personal productivity and organisational productivity are not the same. A marketer can save an hour while the team still wastes the week.
To get a real return from AI, marketing teams need to stop asking how they can make people do the same tasks faster. Instead, ask how AI can redesign the marketing system so human effort moves to higher-value work.
The Hobbs Clock problem
One of the useful ways to understand AI adoption is to separate impressive output from useful output.
For example, it is now possible to use AI coding tools to build a small app in a day, complete with APIs, local servers, device tracking, animations and other details that would once have required weeks or months of specialist effort. That is extraordinary productivity. It is also possible for that app to do nothing important for the business.
This is the Hobbs meter problem. AI makes it easier to build things that feel clever, delightful and technically impressive, even when they are not connected to a real commercial need. That kind of experimentation has value when it builds skill and confidence. It becomes a problem when it is mistaken for strategy.
AI is still waiting for its factory redesign
AI has a familiar promise attached to it. Unlimited productivity always seems to be just around the corner.
Fusion power gives us a useful analogy. In 2022, the US National Ignition Facility achieved a breakthrough by getting more energy out at the target than the laser energy delivered. That was genuinely exciting, but the wider plant still needed far more energy from the grid to power the system. The return at the target looked very different from the return across the whole plant.
AI has a similar gap between local productivity and whole-system impact. Research suggests many people feel AI makes them personally more productive, while far fewer companies can see that productivity reflected in the bottom line. Activity at the target does not always mean gain across the plant.
That does not make AI empty hype. It means the real gains take time, because the systems around it have not caught up yet.
Electricity took decades to transform factories because early adopters often attached electric motors to old steam-era layouts. The big productivity gains arrived when factories were redesigned around the new technology, with one motor per machine, better workflows, new layouts and new operating models.
Marketing is in the same position. Using AI to write the same blog post faster is like attaching an electric motor to an old line shaft. It may improve one task, but it does not redesign the factory.
Marketing budgets are flat, but expectations are not
Budgets are not growing fast enough to absorb another wave of tools, training and experimentation without consequences.
Gartner’s CMO survey found that marketing budgets were 7.7 percent of revenue last year and 7.8 percent this year, which is essentially flat. At the same time, 56 percent of CMOs say they do not have the budget to deliver their strategy, 70 percent want to be AI leaders and only 30 percent say their organisations are ready to scale AI.
There is another squeeze. Around 15 percent of marketing budgets are already being spent on AI, which means many teams are funding AI use from a budget that has not meaningfully increased.
That creates a clear commercial test. If AI is now taking a visible share of the marketing budget, is it earning that investment back? If it is not improving outcomes, reducing meaningful drag or increasing the quality of work, it risks becoming another cost line.
Efficiency is not the prize
At Articulate, our AI vision is built around a simple principle. We want AI to make us better, not merely cheaper.
Efficiency matters, of course. Saving time is useful, and no one should romanticise repetitive admin. But efficiency is the entry ticket, rather than the prize. The more valuable goal is better work, better consistency, better judgement, better systems and better outcomes for clients.
The second principle is human control. We want hands on the wheel, with human judgement, governance, expertise and experience embedded wherever they add value. AI can accelerate tasks, process information and help build systems, but it should not replace the parts of marketing that require taste, strategy, empathy and accountability.
The returns are hiding in workflow redesign
McKinsey’s AI high performers offer a useful clue. Nearly three quarters had fundamentally redesigned workflows around AI, while only around a quarter of lower and medium performers had done the same.
That is the difference between buying a motor and redesigning the factory.
For marketing teams, the value tends to hide in three places.
- The assembly line: Repetitive production work that happens week after week, such as campaign setup, email builds, webinar assets, reporting and CRM admin.
- The knowledge base: Judgement, playbooks, brand voice and methodology that need to become easier for people and AI tools to access and apply.
- The tool layer: Systems where deterministic code, AI synthesis and human expertise each do the work they are best suited to.
There is also a fourth category, which is experimentation. Playful prototypes and speculative builds can be useful for learning, but they should not be confused with operational systems unless they have a clear business purpose, owner and measure of success.
Start with the assembly line
Every marketing team has work that happens repeatedly. Landing pages, email builds, webinar assets, slide decks, reports, briefs, campaign setup, data cleaning, social snippets, newsletter formatting and CRM admin all contain necessary work that does not always deserve human attention from scratch.
At Articulate, webinar production was an obvious candidate. Every webinar needs emails, a landing page, a thank-you page and CTA graphics. Historically, building those assets involved around four hours of careful HubSpot work.
Now the team can add the webinar title, email text and other details to a Notion database, then use an AI-assisted build process to create the assets in HubSpot. The final tool is deterministic code rather than generative AI making decisions on the fly, but AI helped build the system.
The time saving is around four hours a week. More importantly, the saved time does not just vanish into a spreadsheet. It lets the team run webinars more frequently and spend more energy on the substance of each session, including the topic, argument, evidence and delivery.
That is how AI makes the budget go further. It removes production drag so human effort moves somewhere more valuable.
Make routine creative work better
Slide production is another good example.
Articulate has developed a process that turns a deck outline into a PowerPoint file, which can then be imported into Pitch for final use. It is not a perfect one-click miracle, but it saves time, reduces manual design errors and improves consistency.
There are also tools for creating illustrations and animations that support webinar ideas. Previously, fully custom design for every webinar deck would have taken too long. Now the team can raise the quality of the output without spending hours moving pixels around in PowerPoint.
This is sometimes described internally as designer-less design. That does not mean design is unimportant. It means the design value has shifted into the system, the visual language, the tool design and the judgement behind what good output looks like.
AI helps apply the system more consistently. Human thinking still shapes the system.
Ask what your team does every week
The assembly-line question is a useful place to start because it avoids the trap of beginning with the tool. Look first at the work.
What happens every week that is boring, time-consuming and repetitive? Where are people copying and pasting between systems? Where do errors creep in because the task is fiddly? Where does production work stop people from spending time on strategy, client conversations, ideas or creative judgement?
Good candidates might include webinar asset creation, landing page cloning, CRM hygiene, image resizing, campaign setup, report formatting, email build checks, dashboard preparation, meeting summaries, social post variants and sales follow-up prompts.
The best early AI projects often look unglamorous, but they can remove recurring nuisances that absorb significant time and attention.
Put your judgement where AI can reach it
The second hiding place for value is organisational knowledge.
Most companies have plenty of expertise, but it is badly distributed. It sits in one senior person’s head, a brand document nobody opens, a sales methodology PDF, an old tone of voice guide, a half-finished intranet or a folder with a name nobody trusts.
AI can only help apply your expertise if it can access it.
At Articulate, we have been testing a system that puts personas, messaging, tone of voice guidance and a 30-chapter writer’s handbook into a version-controlled GitHub repository. That library can then be accessed through an MCP interface and used inside Claude, ChatGPT and other systems.
In practice, this means brand voice, writing standards and accumulated judgement can be made available to everyone in the business through a simple tool. People do not have to build their own prompts, projects or workarounds because the knowledge arrives where they are already working.
That turns brand voice into something closer to software. It becomes version controlled, maintained, improved and deployed.
Make methods usable, not just documented
The same principle applies to sales and marketing methods.
Articulate has built a sales support tool using Claude projects, HubSpot data, emails, meeting recordings, transcripts and methodology from frameworks such as MEDDPICC and Blair Enns. As a sales engagement develops, each conversation becomes a richer record of the opportunity.
The tool can help analyse next steps, suggest how to move a conversation on and generate an agenda for the weekly sales meeting. The important change is that the methodology is no longer passive.
It is not sitting in a document waiting to be remembered. It is helping the team apply the method at the point of use.
That is the lesson for marketing too. Brand values, messaging frameworks, sales processes and campaign playbooks are only useful if people actually use them. AI gives you a way to bring that knowledge into the workflow, where it can improve consistency and decision-making.
The aim is not to replace judgement. It is to make good judgement easier to apply.
Blend code, AI and human expertise
The third area of value is where deterministic tools and AI tools work together.
At Articulate, we’ve built audit tools for HubSpot portals and websites. The HubSpot tool runs around 200 checks, some through the API and some through a human review interface because not everything can be assessed automatically. It can analyse things that would be tedious to do manually, such as email performance patterns or the number of contacts that have not been emailed in the last year.
The website audit tool runs around 150 checks and draws on sources such as Sitebulb and Google Search Console. Some individual checks would take hours if a person had to do them from scratch.
The key point is that the checks themselves are programmatic, version controlled and repeatable. AI helped build the tools, but the probes are code. AI then helps interpret the output, turning large volumes of structured results into a draft report that identifies patterns, priorities and explanations.
Human experts still review the findings, interpret the implications and help the client decide what to do next. Code handles repeatability, AI handles synthesis and humans handle judgement.
Move human effort to the right place
The audit tools show what good AI-supported work can look like. The human effort has not disappeared. It has moved.
Instead of spending hours manually gathering data, people spend more time maintaining the knowledge base, improving the checks, interpreting the results, communicating with clients and prioritising action. That is a better use of expertise.
It also makes the work more consistent, comparable and repeatable. If the same audit can be run again later, you can compare results over time and see what has changed.
This is where AI can make a budget go further without reducing the quality of the work. The system does more of the heavy lifting, while the human contribution becomes more valuable.
Do not build what you can buy unless it matters
AI makes it much easier to build tools, which also makes it much easier to build tools you should not build.
Before creating your own software, ask whether it will actually move the needle. Is the process unique to your business? Does the tool embed your expertise? Will it create a meaningful advantage? Will it be maintained properly? Is there already a mature product that does the job well enough?
Sometimes the right answer is to build. Sometimes the right answer is to buy. Sometimes the right answer is to prototype something quickly, learn from it and stop.
Not every experiment deserves to become a product.
Watch out for work slop
Giving everyone access to AI does not automatically make them AI engineers. It may simply help them do what they were already doing faster, including things they should not be doing at all.
BetterUp and Stanford have described a problem called workslop. This is AI-generated work created without enough thought, context or concision, which then creates more work for the people receiving it.
One person’s efficient use of AI can become another person’s lost afternoon.
The root cause is often weak leadership. In the study referenced in the webinar, 41 percent of respondents said leaders encouraged them to use AI without explaining how or why.
At Articulate, the position is simple. We reject mindless automation, AI slop and anything our mothers would be ashamed of. A useful rule follows from that. Start with the business problem, not the tool.
Ask what you are trying to improve, how you will know it worked and who benefits. If you cannot say what changes for a client, a colleague or the CFO, do not build it as a business tool.
Treat AI tools as software
Experimentation matters. We are in a fast-moving period and people need space to learn, test and build new skills. Some of the best AI ideas will come from people already experimenting quietly in the background, and those are often the people worth supporting with time, resources and permission.
But undirected experimentation across a whole business is expensive. Once an AI experiment becomes something the organisation might rely on, it needs to be treated like software.
That means a named owner, success criteria, version control, change management, access control, bug reporting, training, deployment, security testing, rollback planning and ongoing maintenance.
None of this disappears because AI helped build it.
High-performing companies are more likely to have a defined way of measuring AI’s impact. That makes sense. If nobody owns a tool, nobody can roll it back and nobody knows whether it works, it is not a business system. It is a hobby.
Change the system, not just the tools
Stafford Beer, a pioneer of cybernetics, had a useful rule. The purpose of a system is what it does.
You may want different results. You may have KPIs, plans and targets that say the organisation should behave differently. But your marketing operations are already optimised to produce the results they currently produce.
If you want different results, you have to change the system.
That is the strategic risk and opportunity with AI. If you drop AI into an unmodified marketing system, you may get the same results faster. You may also get more noise, more inconsistency and more half-built internal tools that nobody maintains.
The technology is not the limiting factor in most marketing use cases. The tools are already good enough for many useful things, and they are improving quickly. The bigger question is whether you are redesigning the work around them.
Redesign the factory
To make AI stretch the marketing budget, don’t start by buying more tools. Start by looking at the design of the work itself. Where does effort get lost? Where do people repeat the same production tasks? Where could AI help the system work better, not just faster?
- Look at the assembly line. Find the work that can be automated or systematised so human attention moves to more valuable tasks.
- Look at your judgement. Work out where your expertise lives and how to make it available inside the tools people actually use.
- Look at your systems. Decide where deterministic code, AI synthesis and human interpretation should each play their part.
Then ask whether every project has a clear purpose, a business case and enough governance to survive beyond the novelty phase.
A practical checklist for AI budget decisions
Before investing in another AI tool, workflow or experiment, ask:
- What business problem are we solving?
- Who benefits if this works?
- Will this improve quality, not just speed?
- Does this remove repetitive production work?
- Does this help us apply judgement more consistently?
- Does this embed our expertise into the workflow?
- Are we building something that should simply be bought?
- Who owns it?
- How will we measure impact?
- What is the kill date if it does not work?
- What needs version control, access control or security testing?
- What human judgement stays in the loop?
- What happens if the tool fails?
- Is this a system or a hobby?
If you cannot answer those questions, pause. Experimentation may still be the right answer, especially while AI is moving so quickly. Just be clear when something is a test, not a transformation project.
Want to make your marketing budget work harder?
Marketing teams are under pressure to do more with budgets that are not growing. AI can help, but only when it is applied to the right problems, with the right systems and the right human judgement around it.
Book a free 30-minute Marketing Strategy Session with Articulate founder Matthew Stibbe. We’ll look at where AI could remove friction, improve quality and help your team get more from the budget you already have.
