Computational Marketing

AirOps or Profound? The wrong question for a technical team: pick the job, then the primitive

Name the failing job first, then pick the tool that does it.

· 8 min read
Cover illustration for “AirOps or Profound? The wrong question for a technical team: pick the job, then the primitive”
Features · November 20, 2026 · 8 min read · 1,764 words

Does the team need to measure where it appears in AI answers, publish governed content, or do both? These are different architectural jobs. A useful evaluation names the failing job first, then chooses the primitive that can do it.

Why the comparison-page genre asks the wrong first question

The comparison-page format starts with two vendors and asks which one wins. That framing assumes they are substitutes before establishing whether they address the same bottleneck. A technical team can therefore spend its evaluation comparing features that solve different problems, then buy both when one would do. What is the job the team cannot currently do?

The genre’s incentive is to produce a verdict. A team’s needs may be less symmetrical than the comparison suggests. It might have a dependable publishing process but no view of AI citations, or detailed visibility data but no reliable route from brief to approved publication. The team needs different things in each case, even if it finds the same vendors in both searches.

The structural shift that created two distinct jobs

AI answer engines can show content regardless of its traditional search results page rank. AI Overviews, AI Mode, Perplexity, ChatGPT, Claude, and Gemini each use their own retrieval and citation logic. A page can earn citations on those surfaces while ranking poorly on Google, and the reverse is also possible. Rank is no longer a sufficient proxy. Tracking keywords alone cannot show whether AI answers cite the team’s content.

That separation creates a production gap as well as a measurement gap. Content shaped around crawl, keyword density, and backlink signals was not necessarily shaped for a language model looking for authority, specificity, and information gain. Earning a citation and ranking on a results page have different success conditions. More publishing does not close the measurement gap. Nor does a better view of citations give the team the approval process it needs to publish content.

The resulting jobs differ in kind. Measurement establishes what engines cite, for which prompts, and what competitors earn. Execution turns research into drafts, review, and published content through a governed pipeline. Most teams have one more pressing than the other, and few have solved both.

What the measurement job actually requires

A team that cannot see where it stands in AI answers has a data-infrastructure problem. Its measurement layer needs to connect prompts to mentions, citations to specific URLs, and those observations to the engines producing them. It also needs citation frequency and competitor share of voice, with enough engine coverage to represent the field the team wants to understand. The team needs to know which pages appear for which questions.

AirOps addresses this through its Insights layer. Analytics tracks share of voice and visibility across AI platforms, while Prompt Tracking shows which queries drive AI visibility. Citation Tracking shows which URLs are cited, along with citation frequency, share, and brand mentions. Access depends on the plan, however. AirOps pricing limits Solo tracking to ChatGPT and places full multi-engine tracking at Pro, covering ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews, Google AI Mode, and more.

A ChatGPT-only view may do the job when the team needs to measure just that engine, but it cannot cover all those engines. Pro requires a substantial monthly investment. For a team that already has a working production process, the question is therefore whether it needs the execution layer at all, or whether a measurement-first tool covers the unsolved job.

What the execution job actually requires

A production bottleneck calls for a governed multi-stage pipeline. The constraint is not AI output quality in isolation, because acceptable text still needs a route through review to publication. A no-code or low-code builder can connect research, drafting, optimisation, and CMS publishing, but the approval path determines what is allowed to ship. Generation is only one stage of the job. Brand grounding, internal knowledge, and direct publishing connections must support that path.

AirOps centres execution on its Workflow primitive. These multi-step pipelines combine LLM calls, web search, data lookups, and content generation, with manual, scheduled, or API triggers. Power Agents provide reusable components for tasks including SERP analysis, brief creation, drafting, metadata, internal linking, and content refresh. Grids uses a spreadsheet-style interface to let the team run pipelines across many rows at once.

Governance adds a different set of requirements. Human Review and Approvals can pause execution and route work to an Inbox, where editors can comment on passages, compare versions, accept or reject changes, and restore earlier content. Brand Kits and Knowledge Bases use tone guidelines, writing samples, and internal data to ground the output. A configurable platform cannot decide who should approve a draft or what the team’s content strategy should be.

Configuration is therefore part of the execution cost. A team must learn to build and direct the workflow. Buying before the workflow is defined increases that burden, because the team is trying to design its process while learning the system used to run it. This configuration burden belongs in the evaluation of how soon the platform can support useful, governed production.

How to audit which job is actually urgent before touching a vendor page

To turn that distinction into a decision, the team should begin with a two-week operating sample. The team must first establish whether it can identify which prompts in its category drive AI mentions and whether its URLs are cited. An inability to answer exposes an unsolved measurement job. Publishing volume does not change that diagnosis. The missing capability is a view of what AI engines are doing with the content.

The second question tests the production path. Can the team take a keyword or prompt through a repeatable workflow to reviewed, published content without manual coordination overhead? If not, execution remains unsolved even when visibility data is strong. The third question isolates approval from production. Are reviewers named, checkpoints explicit, and earlier versions recoverable before AI-generated content publishes?

If the approval path is undefined, adding production capacity first creates a liability. Most teams will find one gap more pronounced than the others. That imbalance identifies the first primitive to evaluate.

A team that finds both jobs unsolved needs to evaluate the connection between them. Does measurement guide a content change that can pass through review and publication, and can the result then be measured? Shared pricing does not establish that connection. The operating sample should contain one visibility gap, one content change, one approval path, and one measured outcome, so the evaluation tests a working handoff.

What AirOps is actually built for and where its architecture constrains it

AirOps is built mainly for execution and also has tools for measurement. Its centre of gravity becomes clear when the earlier requirements are applied to its publishing connections and configuration choices. Direct CMS publishing supports Webflow, WordPress, Contentful, Drupal, Sanity, Prismic, and Shopify. Those connections serve the production handoff. Their relevance depends on whether the team’s actual bottleneck lies between a reviewed draft and its destination.

AirOps runs an orchestration layer over OpenAI, Anthropic, and Google models, allowing model selection by use case and support for teams bringing their own API keys. Solo requires a separate add-on arrangement for that access through contact with AirOps. This is another plan-dependent design choice. Model flexibility and the commercial terms for using it need to be evaluated together.

AirOps pricing allocates Solo 3 knowledge bases and 1 Brand Kit, Pro 5 knowledge bases, and Enterprise unlimited allocations. Free, Solo, and Pro all use task-based billing, so monthly cost is hard to predict without a record of past usage. Workflow design therefore affects the operating estimate. The team needs to weigh the work of setting up the platform and its billing model as part of the production architecture before it chooses a tool.

Taken together, these choices describe a strong fit for a team with an established content strategy and a defined workflow that needs to scale governed production. The measurement layer has the plan limits described earlier, so the lower-priced plans are a poorer fit for teams whose main need is broad visibility intelligence. Neither conclusion requires a universal verdict. The fit turns on which job the team needs to solve and which constraints it can accommodate.

Why the job-first frame also changes what a measurement-first architecture looks like

A measurement-first architecture gives the data layer priority and leaves production optional or composable. It needs daily or near-real-time tracking across multiple engines, prompt-level detail, competitor share of voice, and citation data tied to individual URLs. An aggregate view of a domain cannot replace the connection between a particular prompt, an engine’s answer, and a cited page.

The same logic changes how the platform should fit into an existing stack. A technical team may already have its CMS, review workflow, and data pipeline. It needs a measurement primitive that can connect through an API or MCP connector, rather than a replacement for every working component. Existing systems are part of the requirements. A measurement purchase should not be judged mainly by workflow depth or CMS integration count, because those features serve the other job.

Revenue attribution and media mix modeling extend the purpose of this data. Teams are starting to be able to link share of voice, citations, and crawler activity to a brand’s broader measurement model, so they can move from tracking visibility to measuring business impact. That is a separate demand from collecting mentions. The team must judge the architecture by what it needs to measure, while recognising that visibility alone gives only part of the picture of business results.

For a team whose audit exposes a measurement bottleneck, the practical criteria are engine coverage, prompt granularity, and API composability. If production is also required, the connection to execution becomes an additional test, with those criteria still in force. The governing principle is an architecture that is composable, instrumentable, and grounded in evidence.

The one move that makes any platform evaluation honest

Before opening a vendor comparison, the team should name the specific outcome it cannot produce and the step where that outcome fails. Unknown citation-driving prompts call for measurement, while an inability to publish governed, on-brand content at the needed cadence calls for execution. Failure in both calls for a system that connects them. That diagnosis produces requirements applicable to any candidate, rather than a verdict confined to the names on a comparison page.

The two-week sample then tests those requirements against the team’s approval path, CMS, and the time its editors have available. A polished checklist cannot establish whether those handoffs work.

Pick the job, then the primitive.

More in Features