Budget for AI Discovery Without Abandoning SEO or Existing Demand

Protect essential acquisition work, fund a bounded learning question and define what would justify continuing, changing or stopping the investment.

Share
A calculator and pen resting on handwritten notes.
Illustrative image. Photo by Aaron Lefler / Unsplash

The new AI-discovery proposal fits the budget only if the client cancels its service-page maintenance and reduces a campaign already bringing suitable enquiries.

That trade-off deserves more scrutiny than a slide predicting the future of search. The question is what the next allocation will help the business learn or improve, and what the client gives up to fund it.

An agency can make room for AI discovery without presenting it as a compulsory replacement for everything already working.

Divide the decision into three budget lines

Maintain: work required to keep the current acquisition path functioning. Examples include correcting outdated service information, fixing a broken enquiry journey and meeting existing delivery commitments.

Improve: changes to a known business weakness. The client may receive unsuitable enquiries because a key page does not explain eligibility, or lose useful prospects because procurement details are missing.

Test: a bounded investigation where the opportunity is plausible but the evidence is incomplete. An AI-discovery pilot belongs here when the agency is still learning which buyer tasks and product surfaces matter.

These are proposed management categories, not prescribed percentages. A business with unresolved access or measurement problems may need to fund maintenance before expanding observation. Another may already have the foundations and capacity to test something new.

Google's current guidance says SEO remains relevant to its generative AI Search features and emphasises useful, distinctive content. Google's AI optimisation guide.

That supports a practical budget discussion: do not discard the foundations while paying for a new label. Keep this Google-specific guidance separate from assumptions about every other engine.

Write the investment hypothesis in ordinary language

Avoid “increase AI visibility across all platforms”. It does not identify the business decision or the limit on spending.

For a hypothetical equipment-rental client, a narrower hypothesis could be:

“Prospective buyers struggle to compare equipment suitability. We will investigate that question on two selected surfaces, improve one relevant page using verified operational information and review the evidence before expanding.”

The budget should include the investigation, practitioner input, writing, approval, implementation and review. Funding only the tool or the report leaves the useful page work unfunded.

State what is excluded: a full-site rewrite, unrestricted production changes, additional markets and open-ended testing. If one of those becomes necessary, return to the allocation decision rather than absorbing it silently.

Make the opportunity cost visible

Suppose the agency has a hypothetical 24 hours available for the pilot. It proposes six hours of research, six of evidence gathering and drafting, six of approved implementation and six of review and reporting. These numbers illustrate a capacity conversation; they are not recommended effort levels.

Ask what would otherwise use those hours. Perhaps an existing landing page needs a service-area correction that sales has already requested. If that correction addresses the same buyer question, incorporate it transparently into the work rather than billing two programmes for the same deliverable.

If the pilot instead displaces an urgent repair, the agency should explain why. The novelty of AI discovery is not enough to justify leaving a known commercial problem unresolved.

Also record client time. An unfunded practitioner interview or an unavailable approver can delay the entire plan even when agency hours remain available.

Decide what evidence would change the allocation

Set three possible decisions before starting:

  • Continue: the buyer question remains commercially relevant, evidence is usable and the team can act on a clear next opportunity.
  • Revise: the original question or surface was poorly chosen, or an access or measurement dependency needs attention first.
  • Stop or defer: the client cannot support the work, a higher-priority need emerges or the investigation produces no sufficiently useful next decision.

Do not make continuation depend on manufacturing a positive chart. A pilot can be informative when it shows that broader monitoring is premature.

Likewise, an implemented page improvement is a delivery result, not proof of incremental leads. Keep the commercial outcome under observation without attributing every later movement to the pilot.

Avoid double-counting the benefit

If the same improved service explanation helps a sales representative, an organic-search visitor and a buyer arriving from an assistant, describe the shared asset once. Do not claim its entire commercial value independently for three channels.

For the budget review, show actual effort, unfinished dependencies, implemented work and the next decision. Explain which expected benefit is still a hypothesis. This gives the client a better basis for reinvestment than a dramatic claim about one channel replacing another.

A sensible first commitment may be small. Create a kloovy account to evaluate one bounded workflow against your agency's delivery need. Include the team's review time and current subscription terms in the decision, and expand only when the evidence justifies it.

Sources

  1. Google: Optimizing your website for generative AI features on Google Search