Imagine a marketing team cutting the time it takes to create campaign assets from days to hours, only to spend the rest of the week chasing approvals, checking product accuracy and rebuilding files for local markets. The AI tool has done its job. The brand is still waiting to reach its customers. That is where much of the frustration around AI in marketing and production begins.
ISBA’s July 2026 survey, based on 200 advertiser responses, captures the distance between adoption and impact. It found that 99% were engaging with generative AI, yet only 14% reported a significant impact on business results. Engaging with a technology clearly tells us very little about how successfully it has become part of the business. [1]
The wider picture is similar. McKinsey’s August 2026 global survey found that 44% of respondents reported AI scaling across their organisations, up from 38% a year earlier. Yet the share reporting a positive contribution to earnings before interest and taxes remained essentially unchanged at 37%. Deployment is progressing faster than reported financial impact. [2]
These surveys measure different things, but they strengthen my view that we need to pay much more attention to what happens around the technology. Production makes this particularly visible because content passes through so many hands before anyone can see it.
A new tool enters a system of briefs, budgets, agency relationships, asset libraries and approval processes. If those arrangements remain unchanged, a faster production task may simply move the queue somewhere else. More generated content can mean more work for the people checking it. Cheaper first drafts can become expensive finished assets once correction, localisation and reviews are included.
Ownership compounds the problem. Marketing may sponsor the ambition, IT approve the software and Procurement negotiate the contract, while producers are left to make everything work together. Each team can meet its own objectives without anyone being accountable for whether the complete process improves.
BCG’s July 2026 CEO research points to precisely this execution problem. Respondents were more likely to cite difficulties linking AI to financial outcomes and redesigning workflows, roles and incentives than technical barriers. Its findings also show how easily experimentation becomes an organisational habit: 64% said their companies pursued AI pilots, while only 26% embedded AI within a broader business transformation. [3]
There is nothing wrong with trying things. Discovery gives creative teams room to learn what emerging technology makes possible. But a successful demonstration needs a route into everyday work. Someone has to own the next stage, fund it and decide what evidence would justify scaling and wider use across the organisation. Otherwise, the organisation accumulates demonstrations while its production problems remain.
The starting point should therefore be a specific problem that people recognise. Perhaps regional teams cannot adapt a campaign before the moment has passed. Perhaps product imagery is repeatedly recreated because finding or reusing approved assets is not possible or to. Perhaps most of a creative team’s time is lost in the multiple hand-offs – manually moving data between disconnected systems – and goes into mechanical versioning. Each problem points towards a different intervention. Generative AI may help; automation, but most of the time a better asset lifecycle management approach or a change in responsibilities may be more useful.
Before choosing the technology, establish what happens today. How long does work take from detecting what’s trending, turning it into a brief and an approved delivery? What does it cost, including revisions? Where are the bottlenecks? What opportunities are lost through delay? Where does quality suffer? This gives the team a baseline against which to judge improvement and helps prevent that the project’s purpose is simply experimenting with and trying out the latest tech or the latest most impressive feature.
Validation should then follow a representative piece of work through the whole process. For a localisation brief, that means using actual brand assets and testing the proposed approach with the people who will produce, review and approve the material. Include difficult formats and exceptions. Measuring not only the generation time but the human effort needed to review and correct the outputs. A result dependent on constant intervention from a specialist is a different proposition from one an ordinary production team can repeat.
Turning validated improvements into a normal practice that can scale, requires decisions that rarely feature in a technology demonstration. Who maintains approved assets? Who can change the workflow? When must a human intervene? How will agency scopes and fees change when responsibilities move? Teams also need time to learn; timely support and a clear understanding of how their expertise will be used and where value is being added. Asking people to transform production on top of their existing workload is unlikely to produce lasting change.
Governance belongs in that design. Rules for permitted materials, product accuracy, approvals and records should be clear before work begins. ISBA’s February 2026 framework for AI-enabled content creation reflects the same concern, connecting AI adoption with consumer trust, human judgement and the wider marketing operating model.[4] When teams understand which decisions they can make and which require escalation, they can streamline processes, reduce unnecessary delays and accelerate production while working with greater confidence.
This becomes more pressing with agentic systems that can take actions across a workflow. A brand needs to establish what a system is authorised to do, which information it can use and who is accountable for its decisions. Automating a process makes clarity about that process more consequential.
Brands also need room to change their minds. Supplier recommendations can reflect the capabilities those suppliers sell, and a platform that suits one task or a specific campaign may be poorly suited to another. Independent evaluation should examine fit, integration effort, full operating costs and the ability to move assets and data if requirements change. Consolidating suppliers can simplify contracting, but shared ownership alone does not resolve operational handoffs.
This is the thinking behind The Fluid Network, the vendor-neutral operating model we have developed at MurphyCobb. It connects creative, production, data and technology capabilities around the brand’s requirements, or that of a specific campaign, trend or given moment, with shared governance and clear accountability. Existing agencies can work alongside specialist studios, independent talent and technology providers. The brand retains ownership of its ecosystem, including the data and knowledge generated through it.
This growing body of knowledge can inform and improve the brand’s AI systems, allowing the organisation to build on what it learns with each project. As work moves through the ecosystem, production and performance data can feed into future briefs, helping teams understand which approaches work best for different audiences, formats and markets.
Making this possible requires common standards, compatible systems and clear responsibility for coordinating delivery and capturing learning. These foundations should also allow the brand to introduce or replace individual capabilities without rebuilding its entire operation. The approach can begin with one priority workflow and expand as the evidence supports it.
That emphasis on turning technology into measurable impact also underpins Amplify – MurphyCobb’s production innovation and transformation business unit. Through a practical, evidence-based approach, Amplify helps brands diagnose production problems, validate emerging capabilities in real conditions, and embed and scale what works, with continued support as their needs evolve.
The definition of success deserves equal care. ISBA found that advertisers prioritising effectiveness, not just efficiency, in their use of generative AI were twice as likely to report significant business impact. That is an association, rather than proof of causation, but it is a useful challenge to programmes judged mainly on output volume and savings. [1]
A faster workflow matters when it helps a brand respond while the opportunity is still relevant. Lower costs matter when the saving blends seamlessly into the full production process. More creative options matter when people have the time and judgement to develop and select the ones worth showing. Teams should measure those operational gains alongside audience response and commercial results, without assuming one automatically causes the other.
I believe technology should amplify creativity and productivity, with craft at the centre. That means giving people better conditions in which to make work that matters to its audience. Closing the adoption-to-impact gap requires leaders to follow an AI initiative beyond technology demonstrations or isolated pilots, into the briefs, decisions, relationships and everyday working practices that determine whether it succeeds. The goal is to embed improvements that deliver measurable value and can be repeated and scaled. That is where a brand earns the return on its investment.
Source notes for the editor
References supporting the draft. Source notes are outside the article word count.
[1] ISBA, Major survey reveals impact of Gen AI on Effective Advertising, 22 July 2026.
Survey based on 200 advertiser responses; supports the engagement, significant business impact and effectiveness findings.
[2] McKinsey, The state of AI in 2026: On the road to ROI, 25 August 2026.
Global survey; supports the 44% versus 38% enterprise scaling comparison and the essentially unchanged 37% reporting positive EBIT contribution.
[3] BCG, CEOs Are Starting to See Value from AI. Now Comes Execution., 22 July 2026.
Supports execution barriers and the 64% pilot and 26% broader transformation figures.
[4] ISBA, New ISBA framework supports responsible AI enabled content creation, 26 February 2026.
Public announcement of the framework; supports the connection between governance, trust, human judgement and operating models.
