Nick QuirkThe interesting thing about AI in SEO isn't what it does to rankings. It's what it does to your P&L.
Every agency owner I talk to has the same three problems right now: deliverables that used to take eight hours now take three (and clients somehow know it), organic traffic charts that dropped 20-40% on informational keywords without a single ranking loss, and a team that's quietly using six different AI tools with no shared standard for what "done" means.
Those are operations problems, not SEO problems. Here's how to handle them.
What AI actually changed about SEO work
Two separate shifts happened at once, and conflating them causes bad decisions.
Shift one: the SERP changed
AI Overviews, AI Mode, and LLM-based answer engines now sit between the query and the click. The measurable effect on client accounts we've seen: informational queries lose 25-45% of clicks even when the ranking position holds. Transactional and commercial-intent queries have barely moved. Local pack results are largely untouched.
This means your reporting narrative has to change before your client notices the traffic dip and calls you. If your monthly report leads with sessions, you're setting yourself up for a conversation you'll lose. Lead with qualified conversions, branded search volume, and assisted revenue.
Shift two: the production cost of SEO deliverables collapsed
A 2,000-word content brief that took a strategist 90 minutes takes 25 with a good prompt and a review pass. A crawl export of 40,000 URLs that used to eat a full day of manual categorisation gets clustered in an hour. Keyword research that filled two days now fills half of one.
This is the shift that breaks agency economics, and it's the one most owners are underreacting to.
The margin trap: why faster work made agencies less profitable
If you bill hourly or bill against a monthly hour allocation, AI just cut your revenue. You do the same work, you invoice for fewer hours, the client is delighted, and you're down 20%.
If you bill fixed-fee retainers, you're temporarily fine — but only until a competitor quotes the same scope 30% cheaper because they've done the same maths and want the logo.
Here's the arithmetic from a real 12-person agency we worked through this with. Their standard $6,500/month retainer included:
- 4 content briefs (was 6 hours total, now 2.5)
- 1 technical audit refresh (was 8 hours, now 5)
- Link prospecting and outreach (was 12 hours, now 9 — outreach doesn't compress much)
- Reporting and the client call (was 5 hours, now 4)
Total: 31 hours down to 20.5. At a $130 blended rate, that's $2,665/month of "freed" capacity per client. Across 14 retainers, that's the equivalent of two full-time strategists appearing out of nowhere.
The agencies that won here did one of two things. They either sold the freed capacity back into the same retainer as new deliverables (digital PR, CRO testing, YouTube/answer-engine optimisation) and held the price, or they took on more clients per strategist and let the margin land. The ones that lost simply invoiced fewer hours and wondered why Q3 was flat.
Repricing: what to charge for now
Stop pricing the artefact. Price the outcome and the accountability.
Nobody pays $400 for a content brief anymore because they know a $20 subscription produces something that looks similar. But clients will pay for a content programme where someone owns the editorial calendar, defends topic selection against actual search data, and is on the hook when it doesn't convert.
Practical version of that shift:
- Move from deliverable counts to programme outcomes. "12 briefs per quarter" becomes "own the organic content programme for these 6 product lines, target 40% growth in non-brand demo requests."
- Charge separately for judgement work. Migration oversight, dev-team liaison, algorithm-recovery diagnostics, and executive stakeholder management are the things AI cannot touch. These should be your highest-rate line items.
- Keep tracking hours internally even after you stop billing them. You need to know which clients are quietly eating 45 hours against a 20-hour scope. Time tracking becomes a profitability instrument, not an invoicing one.
That last point matters more than it sounds. Once you decouple hours from invoices, most agencies stop tracking altogether — and lose the only early warning system they have for scope creep.
Rebuilding your workflow around AI-assisted work
The failure mode I see constantly: an agency adopts AI at the individual level. Three strategists each build their own prompt library, quality varies wildly between accounts, and nothing is transferable when someone leaves.
Fix it at the process level instead.
Standardise the prompt, not the person
Every recurring deliverable should have a documented, versioned prompt stored with the project template — not in someone's personal notes app. A content brief prompt should include your house style, the client's ICP, banned phrases, internal linking rules, and the exact output structure your writers expect.
When that lives in the workspace next to the task, a new hire produces on-brand output in week one instead of week six. This is where AI project scoping pays off too — scoping a new client engagement from a discovery transcript takes 20 minutes instead of an afternoon, and the estimate is more consistent across your team because everyone starts from the same structure.
Make QA a named task with an owner
AI output fails in specific, predictable ways: invented statistics, hallucinated internal URLs, confident claims about a client's own product that are wrong, and a flattening of tone toward generic B2B mush.
Add a QA step to every AI-assisted deliverable in your project template. Not a checklist someone reads — a task with a name on it and 20 minutes allocated. If it isn't a task, it doesn't happen under deadline.
Connect the audit output to actual work
AI has made audits trivially fast to produce and no easier to execute. A crawler plus an LLM will hand you a 60-page audit in an hour. That document is worthless until it becomes assigned, prioritised, sequenced tasks with dev dependencies mapped. The bottleneck moved from analysis to execution — which is exactly why turning technical audits into actionable to-dos is now the higher-value half of the job.
Track algorithm and AI-surface volatility next to delivery
With AI Overviews expanding and core updates landing more frequently, "did the update cause this or did we?" is a monthly question. Keep algorithm update tracking alongside client deliverables so you can overlay a traffic drop against what your team shipped that week. It turns a defensive client call into a diagnostic one.
The tooling question
Generic project management tools were built for software teams and adapted for agencies. That mismatch got worse with AI, because the new bottlenecks are agency-specific: prompt libraries tied to client context, retainer hour burn against fixed scopes, and search data sitting next to the tasks it justifies.
You can bolt this onto Asana or ClickUp with enough custom fields and automations — plenty of agencies do. But maintenance cost is real, and someone owns it. If you're evaluating the trade-off, the agency PM tool roundup covers where each option breaks down at scale.
Whatever you choose, the requirement is the same: one place where the scope, the hours, the search data, and the AI-assisted output all live. Splitting those across four tools is how you lose 15% margin without ever seeing it on a report. PeakKR was built specifically for that overlap, which is why we're opinionated about it.
What to do in the next 30 days
- Audit one retainer end to end: log actual hours per deliverable and compare against what you scoped 12 months ago
- Calculate your freed capacity across all retainers — then decide explicitly whether you're selling it back or banking it
- Move every recurring deliverable's prompt out of individual accounts and into a shared, versioned project template
- Add a named QA task with a real time allocation to every AI-assisted deliverable
- Rewrite your monthly report template to lead with conversions and branded search, not sessions
- Segment client keywords by intent and quantify informational-query exposure before the client asks
- Reprice at least one renewal around programme outcomes rather than deliverable counts
- Keep internal time tracking running even where you've stopped billing hourly — it's your scope-creep alarm
