What AI Should Change In Your Agency Relationship
Many companies now assume that agency work should cost less because artificial intelligence can produce a first draft, summarise a document or generate several campaign ideas within seconds. The assumption is understandable. Some tasks have become materially faster, and clients should expect agencies to pass those efficiencies on.
The difficulty begins when faster production is treated as evidence that the entire service has become simpler.
For communications and marketing teams, AI changes more than the amount of time required to write a text. It changes which parts of agency work remain valuable, how assignments should be scoped and what clients should examine before comparing fees. The most useful question is no longer whether an agency uses AI. Most do. Companies need to understand what the technology has improved, which responsibilities still require experienced judgement and whether the agency is using the additional capacity to strengthen the work or merely to produce more of it.
Expect routine work to become faster
AI has already reduced the time required for a range of operational tasks. Agencies can transcribe interviews, summarise background material, compare documents, develop initial structures, produce basic adaptations and prepare first versions more quickly than before.
Clients are entitled to benefit from that productivity. Routine work should not continue to be priced as though every stage were performed manually. An agency that uses AI internally while preserving outdated production estimates without explanation will struggle to justify its commercial model.
The benefit does not always have to appear as a direct reduction in fees. It may take the form of shorter delivery times, more thorough research, additional language versions, stronger preparation for workshops or a broader set of options within the same budget. What matters is that the efficiency becomes visible somewhere in the mandate.
Clients should ask where AI is being used, which steps now require less time and how the saved capacity is being redeployed. A vague claim that technology has been “integrated into the workflow” says very little. A credible agency should be able to explain the operational effect.
Do not confuse the first draft with the finished work
Generative AI has lowered the cost of producing plausible language. It has not removed the work required to determine whether that language is accurate, distinctive or appropriate for the organisation publishing it.
A first draft can be created quickly. It may still contain an unsupported assertion, a dated figure, an invented source, an exaggerated claim or wording that does not reflect the company’s actual position. It may sound professional while remaining strategically weak.
For clients, this distinction matters because much of the visible output can now be generated before the more demanding work begins. The time saved in drafting should create room for stronger editorial control, factual verification, stakeholder input and senior review. Where that does not happen, AI may simply accelerate the production of material that the client must later correct.
The quality of an agency relationship should therefore be assessed by more than speed. Clients should examine whether the agency challenges weak assumptions, checks the evidence, recognises reputational risk and understands when a technically correct text still sounds wrong for the organisation.
The faster the first version becomes, the more important the review process becomes.
Swiss companies need more than fluent translation
The Swiss market makes superficial efficiency especially risky. A company may need to communicate across German-, French- and Italian-speaking regions, often while maintaining one institutional position. AI can produce multilingual copy rapidly, but linguistic accuracy does not guarantee equivalent meaning.
A message that feels appropriately restrained in German may appear distant in French. A direct translation may use terminology that is technically correct but unusual in a particular region or industry. Corporate language that works in Zurich may require adjustment before it is credible in Geneva or Lugano.
Internationally active Swiss companies face an additional layer of complexity. Communications may need to satisfy Swiss expectations while remaining suitable for European, British, American or Asian audiences. The legal and reputational implications of a claim can change between markets, even when the wording appears similar.
Clients should therefore distinguish between translation and localisation. AI is highly effective at producing an initial version. Market-sensitive review still requires people who understand the professional vocabulary, cultural expectations and institutional context of the audience concerned.
This is one area in which a lower production cost should not be achieved by removing expert review. The result may be cheaper, but it is also less reliable.
AI governance is part of the service
Companies should also understand what happens to their information when an agency uses external AI systems.
Interview transcripts, strategic documents, employee cases, unpublished financial information, client data and internal correspondence may contain material that should not be entered into a public model. The fact that a tool is widely used does not make every use appropriate.
An agency should be able to explain which systems are approved, what information may be processed, whether data is retained, who has access to it and where human review is mandatory. For regulated organisations, this should form part of supplier assessment rather than being left to informal reassurance.
Swiss companies operating in financial services, insurance, healthcare, pharmaceuticals or other sensitive sectors should expect a particularly clear answer. The same applies to businesses whose activities bring them within the scope of European regulation or client-specific confidentiality obligations.
AI policy is not an internal technical detail. It is part of how the agency manages the client’s risk.
More output is not automatically more value
One of the most common effects of AI is an increase in content volume. Agencies can produce more articles, posts, campaign variations and executive drafts without a corresponding increase in headcount.
That may be useful, but it can also create an organisation that publishes more and decides less clearly.
Marketing and communications teams already face crowded approval processes. More drafts mean more material for specialists, executives, legal teams and compliance functions to review. When the additional output is only marginally different, the supposed efficiency is transferred to the client as a larger editorial burden.
The relevant measure is therefore not how much content an agency can now produce. It is whether the agency can improve the selection of what deserves to be produced.
Clients should expect agencies to use AI to widen the field of possible ideas and then narrow it intelligently. Ten weak concepts are not more valuable than three well-developed ones. A large editorial calendar is not evidence of strategy. An agency that recommends publishing less may be providing better advice than one that fills every available channel.
The commercial value lies in relevance, timing and authority, not in volume alone.
Agency briefs should change as well
Clients influence the quality of AI-supported work through the way they commission it.
A brief that asks for a fixed number of articles, posts and newsletters encourages the agency to optimise production. A brief that asks how the company can establish authority around a business priority creates room for research, audience analysis, expert input and more selective content choices.
This does not mean abandoning deliverables. It means connecting them to a clearer purpose.
Companies should be explicit about the decision or commercial objective behind the communication. Who needs to understand something differently? What evidence does the organisation possess? Which claims can it support? Which internal experts should contribute? What should the audience do after seeing the material?
AI can accelerate parts of that process, but it cannot resolve a weak strategic premise. When the brief lacks a clear point of view, the technology tends to produce smooth, generic communication because generic communication is the easiest output to generate.
A stronger client brief makes it easier to see whether the agency is contributing judgement or merely execution.
Pricing will need to become more transparent
The traditional hourly model becomes harder to interpret when routine production takes less time. Clients should expect greater clarity about what they are paying for.
Some tasks should become cheaper. Straightforward adaptations, initial research summaries, transcription and basic drafting no longer require the same level of manual effort. Agencies that continue to price these activities as scarce specialist work should be challenged.
Other parts of the mandate may retain or increase their value. Senior advice, original research, access to specialists, regulatory interpretation, multilingual review, crisis preparation and final accountability are not interchangeable with automated output.
The client’s task is to separate these elements rather than applying one assumption to the whole fee.
A lower overall price can sometimes be achieved by reducing senior involvement, limiting verification or moving more review work back to the client. This may be a rational choice for low-risk material. It is less appropriate when the subject affects the company’s reputation, market position or regulatory exposure.
A useful proposal should make the trade-offs visible. Clients should be able to see which work is automated, which is reviewed by experienced staff and where responsibility for the final recommendation sits.
The right questions have changed
When selecting or reviewing an agency, companies should ask more specific questions than whether AI is used.
They should ask which parts of the workflow have become faster, how confidential information is protected, how facts are verified and who reviews multilingual material. They should understand when senior consultants become involved and whether the agency is prepared to advise against unnecessary communication.
They should also ask what has improved for the client. Has the research become deeper? Are recommendations better supported? Is delivery faster? Has the agency created more time for interviews, analysis and strategic discussion? Or has AI simply increased the number of drafts entering the approval process?
The answer will reveal whether the technology has strengthened the service or merely reduced the agency’s internal workload.
Clients should expect a better service, not just a cheaper one
Artificial intelligence has made some agency tasks significantly more efficient. Companies should not pay for avoidable manual work, and agencies should be transparent about where technology has changed the effort required.
The larger opportunity is to use those efficiencies to improve the quality of the mandate. Faster research can support better preparation. Faster drafting can create more time for verification. Easier adaptation can allow communication to be shaped more carefully for different Swiss regions and international markets.
This only works when clients know what to demand.
AI has reduced the effort required to generate content. It has increased the importance of deciding what deserves to be communicated, how far a claim can be defended and who accepts responsibility for the final result. Companies that understand this distinction will negotiate better agency relationships and recognise more quickly where genuine value is being created.
