AI made content cheaper. It also made generic marketing easier to ignore

September 22, 2026

Producing another article is a weak answer to a buyer who is questioning whether the existing content is useful. For vendors selling content platforms, AI services or marketing support, the September US marketing roundtable on content strategy presents a more demanding brief: help the organisation say something its audience has a reason to read.

The discussion included disappointment with recycled AI material and an account of moving towards case studies and expert interviews after an initial period of success with AI-generated articles. It also included practical uses of AI to repurpose existing material. The evidence therefore supports a selective approach to enterprise AI content strategy. It does not justify a blanket claim that AI content fails or that human production guarantees performance.

The commercial question is what the technology helps the organisation contribute. If the input consists of generic material, a polished output may still give a prospective buyer little basis for judging the organisation’s expertise. Vendors should explain how their offer helps capture useful knowledge, develop it responsibly and put it into a form that supports a real audience decision.

What the roundtable says about enterprise AI content strategy

September’s content discussion describes a move towards human contributions, including subject matter expertise and interviews. The reported concern was differentiation: material built from recycled information was becoming less effective for the organisations discussing it. One account connected a change in content approach with declining search placements, but the summary does not establish AI generation as the sole cause of that decline.

That distinction matters for vendors. A pitch based on fear that all AI-generated content will lose visibility would go beyond the source. A more defensible proposition is to help the buyer understand where its content lacks expertise, relevance or an identifiable contribution, and then improve the production process around those weaknesses.

The same session records experimentation with turning long broadcast material into short social clips. Here, AI was being applied to material the organisation already possessed, with ongoing manual adjustments and testing of cadence. The contribution existed before the tool repackaged it. This gives vendors a useful starting point: identify the organisation’s valuable source material before proposing a higher output target.

The Leadership Board’s interpretation is that these buyers were wrestling with content effectiveness as well as production efficiency. A vendor should be prepared to discuss both. A faster workflow has value, but the proposal needs to explain how the work produced will remain worth publishing.

Start with knowledge the buyer can stand behind

The July executive AI discussion raised a related concern: AI could produce similar recommendations for competing organisations. Contributions considered the role of proprietary context in making outputs more useful. This does not mean that any confidential information should be fed into a tool. It means vendors should discuss what approved knowledge can support a distinctive and accurate contribution.

For a content engagement, a sensible discovery exercise would identify existing research, product expertise, customer questions and approved examples. These are recommended inputs rather than a claim that every organisation in the roundtables maintained the same resources. Ask who owns the material, whether it is current and what permission exists to use it externally.

An expert interview should have a defined purpose. It could explain how a particular problem arises, where implementation becomes difficult or what conditions affect a result. Those details give an editor something substantive to develop. A transcript alone still needs judgement about what belongs in the final piece and what requires clarification.

Vendors can make this process easier by showing how expert contributions enter the workflow and how their meaning is preserved. The demonstration should include a correction or an unresolved question, because real expertise is not always a neat set of publishable answers. A system that surfaces uncertainty for review can be more useful than one that confidently fills the gap.

Make human expertise practical to use

Calling for more human content leaves a practical problem unresolved: specialists have other work to do. A vendor proposal should show how the organisation can contribute expertise without asking every expert to become a full-time writer. The source supports the importance of those contributions; the operating approach needs to be designed for the buyer.

One recommended model is a focused brief followed by an interview, an edited draft and a clearly scoped factual review. The expert’s task is to verify substance and resolve questions. The editorial team’s task is to make the piece readable and useful. AI may assist with organising or adapting approved material, provided the organisation’s rules allow it.

The hand-offs deserve attention. If reviewers receive an unstructured draft with no indication of what changed, the apparent saving in writing time may simply become additional review work. A vendor should explain how comments, source references and approval decisions remain visible. That is a more meaningful demonstration than generating a finished-looking article from a single prompt.

This also creates an opportunity for services alongside software. A buyer may have a capable tool but lack a consistent interview process or enough editorial support. Diagnose the missing capability before recommending more automation. The right scope should address the work preventing useful expertise from reaching the audience.

Define quality before increasing volume

The September business impact session includes an account of increased content output while maintaining quality. The content strategy session, meanwhile, records concern about the effectiveness of AI-heavy material. These experiences should sit together. They show that a production increase and an editorial problem can both exist within the source set, without establishing a universal rule about either.

Vendors should ask the buyer to define what acceptable content must do. A practical review might consider factual support, relevance to the intended reader, clarity and the contribution beyond existing material. The criteria should reflect the organisation’s purpose. An explanation of a complex service may need different checks from a short adaptation of approved event footage.

The proposal should also specify what happens when a draft does not meet those criteria. Does an editor revise it, request more evidence or decide against publication? Include that work when estimating efficiency. A content system that produces many drafts can create a burden if the buyer cannot identify which deserve attention.

For the sales demonstration, use a realistic brief and show the review process. Invite the buyer to inspect a weak or incomplete output as well as a strong one. This reveals how the product supports editorial judgement and gives the buyer a more credible basis for estimating the effort required after purchase.

Repurpose material with a reason for each format

The September example of adapting broadcast material into social clips offers a useful distinction between generating more material and finding another use for existing material. The source describes continued experimentation with frequency and process, including manual adjustment. It does not provide a fixed publishing cadence that other organisations should adopt.

A vendor should therefore ask what each adaptation is meant to help the audience do. A short extract could introduce a useful question, while a longer article explains the conditions behind the answer. Both formats should retain enough context to avoid turning a qualified observation into an absolute claim. Repurposing should preserve meaning as well as brand presentation.

This becomes particularly important with interviews and customer examples. A concise quotation may lose the circumstances that made the original statement accurate. The workflow should provide a way to check the extract against its source and return it for review where necessary. That is a concrete requirement a buyer can evaluate.

Distribution also needs editorial judgement. The roundtable’s experimentation with social clips suggests that format and cadence require observation, rather than a universal automation setting. Vendors should offer a way to learn from the buyer’s own results and adjust the approach. A large library of variations is useful only if the team can decide how to use it.

Measure whether the content helps the audience

September’s content discussion questioned familiar performance measures and considered time spent on page, conversion and citations. It also described erratic analytics and concern about bot activity. These are reported measurement challenges. They should not be rewritten as a claim that any particular metric has become universally obsolete.

For vendors, the recommendation is to agree the purpose of each content type and choose evidence that bears on that purpose. A technical explanation may support a buyer’s understanding before an enquiry. A customer story may help an internal sponsor explain a use case. A short clip may introduce an audience to a topic. Counting assets does not establish whether those purposes were served.

Where the buyer uses engagement measures, explain what they can and cannot show. Time on a page may be one useful observation, but it needs context. A conversion measure should identify the action being counted. If analytics are affected by unreliable traffic, the measurement plan should acknowledge that limitation instead of presenting a clean-looking report as certainty.

Keep production measures alongside audience measures. The buyer needs to know both whether the workflow is manageable and whether the resulting material supports its objectives. This creates a balanced evaluation of an AI content platform: the work becomes easier to produce, and there is a defined method for assessing its usefulness.

Treat declining traffic as a question to investigate

The source discussion links content strategy with changes in AI-mediated discovery and fewer direct visits. It also raises the possibility that information delivered away from a website may still support later engagement. These are issues the participants were exploring, rather than evidence of a settled replacement for website measurement.

A vendor should avoid explaining every fall in traffic as either poor content or successful AI visibility. Both conclusions require evidence the summaries do not provide. Instead, help the buyer separate the questions: is the content useful, can the intended audience find it, and what can the organisation observe about the resulting journey?

This distinction helps keep an enterprise content strategy coherent. Improving an article’s substance is different from improving how it is made available, although both may be necessary. A proposal should state which problem the vendor is addressing and what evidence will support the work. If another team controls the website or analytics, include that dependency.

The July discussion of frictionless experiences adds a related point: some teams were reconsidering complex lead capture and placing more emphasis on accessible content before requesting details. For vendors, that is a prompt to examine the reader’s experience around the content, rather than judging effectiveness solely by the number of form submissions attached to it.

Preserve the brand without flattening the voice

September’s executive AI session describes work on shared brand context and concerns about unauthorised tools producing off-brand material. That evidence connects content quality with governance. An enterprise buyer may need consistency across contributors while still allowing specialists to explain a subject in a credible, natural voice.

Vendors should distinguish rules that protect accuracy or identity from preferences that can be applied with editorial judgement. A factual claim may require verification. A description of a product may need approved terminology. An interview may retain the expert’s individual phrasing where it helps the reader understand the point. The source does not prescribe a universal brand workflow, so the design should be agreed with the buyer.

Show how the system uses approved context and how a reviewer can challenge its application. If every draft is pushed into the same promotional style, the organisation may lose the distinctiveness it sought from expert contributions. The buyer should be able to examine what the tool changed and decide whether the change serves the piece.

Governance also affects the source material itself. Confirm that examples can be used, that names have the necessary approval and that confidential details are excluded. In this series, roundtable contributions are deliberately anonymised. The wider vendor lesson is to make permissions part of content development from the start.

Reframe the offer around useful content

An enterprise AI content proposal should make its editorial promise concrete. Explain where the knowledge will come from, who will verify it, how it will be adapted and what evidence will help the buyer judge the result. Include the review effort in the scope, and make clear which responsibilities remain with the organisation.

For a pilot, choose a bounded subject where the buyer has approved expertise and a clear audience need. Develop the material through the proposed workflow, record the effort involved and examine the response using agreed measures. This is a recommended way to evaluate the offer; the roundtables do not provide a standard pilot design or a guaranteed outcome.

The September discussion gives vendors a valuable opening. Content buyers in these conversations were already asking about authenticity, expertise and effectiveness. A proposal that addresses those questions can move beyond a contest over production speed. It can show how the organisation will use AI while maintaining a reason for its audience to pay attention.