Every marketing team has a strategy. Far fewer have a production system. The gap between a well-conceived campaign and a live one is where most teams lose time, quality, and credibility, and the causes are almost always mechanical: intake forms that capture the wrong information, naming conventions that get ignored until reporting is a mess, QA checklists nobody updates, approval workflows that add days without adding value.
Knak’s 2026 Marketing Production in the Age of AI report, a survey of 333 enterprise marketing decision-makers, puts numbers on the damage: 82% of teams spend at least half their time on production instead of strategy, and 85% missed a campaign launch date in the past 12 months because of workflow constraints. We pulled campaign production lessons from 207 practitioner conversations on Humans of Martech, from Wyatt Bales building fully automated brief-to-launch pipelines, to Abby Gailey’s negative QA checklists, to Angela Rueda scaling Knak across Meta’s global marketing organization, to assemble the playbook for a campaign production system that ships reliably and scales without adding headcount.
Why campaign velocity is the metric most MOps teams aren’t tracking
Most marketing operations teams measure campaign performance. Almost none measure campaign production performance: how long it actually takes to go from brief to launch, how many people it takes to get one out the door, and where the time disappears in between. The result is a hidden constraint on how much any team can learn in a quarter.
Josh Kim, Notion’s Growth Marketing Lead, made the point in growth terms: if you can only ship 10 experiments in a quarter and historically 1 or 2 turn into wins, your impact is capped by shipping velocity. Every blocked handoff costs you a shot, whether it’s a data engineer who can’t get to a new segment build or a QA backlog that takes 2 weeks to clear. In Episode 110, Josh described each bottleneck as “a failure point” that reduces the total number of shots your team gets.
Wyatt Bales, who spent years building automated campaign infrastructure in banking and healthcare, took that logic further. His team pushed every step of the production process, from intake to delivery, into Tableau, so executives could see how efficiently campaigns were produced alongside how they performed: how many people it took to launch one, and how long it took. In Episode 80, he noted that this visibility matters most in industries that struggle with staff retention, because it shows you which processes are disproportionately labor-intensive before someone leaves and takes that knowledge with them.
“For growth teams, shipping velocity is like the most important ingredient to our success. Because in a quarter time, if you can only get 10 shots or experiments up, and you can count on maybe one or two of those being wins, then making actual impact is heavily dependent on shipping velocity.”
Josh Kim, Episode 110
Angela Vega, Expedia’s martech leader, made the cultural argument in Episode 192: real leadership in marketing ops comes from the unglamorous work of turning ideas into live campaigns. Execution turns collaboration into credibility. Every project you push across the finish line strengthens your team’s standing with stakeholders, and stakeholders remember which team always has campaigns live and which team always has campaigns in review. If you want to know what your production system is costing you, measure the calendar. Count weeks from brief submission to go-live. You’ll find the bottlenecks faster than any process audit will.
Knak’s report gives you numbers to measure against. The 82% of time going to production covers building, formatting, reviewing, and sending, which leaves 18% for strategy and planning. That split holds across every team structure in the survey, whether hybrid, centralized, or decentralized, which puts the problem in the workflow rather than the org chart. Among the 85% of teams that missed a launch date, 10% missed one more than 5 times. The 3 biggest culprits were consistent across company size and industry: getting approvals and sign-off at 47%, design and creative production at 38%, and coordinating across multiple teams and stakeholders at 36%.
The cost side is what gets a CFO’s attention. 36% of teams need at least one full business day to produce a single email from brief to send, and 60% involve at least 4 people. Applying BLS wage data for marketing managers, Knak prices a full day of 4 people at roughly $2,400 per email. Even the best case, an hour of work with 4 people in the room, runs past $300 in internal labor per send. Multiply that by your monthly send volume and campaign velocity stops being an ops complaint and becomes a budget line.
| Time from brief to send | Share of teams |
|---|---|
| Under 1 hour | 13% |
| 1 to 4 hours | 29% |
| 4 hours to 1 day | 22% |
| 1 to 2 days | 17% |
| 3 to 4 days | 11% |
| 5 to 10 days | 6% |
| 10+ days | 2% |
How to structure your campaign intake so the brief captures what you actually need
The brief is the contract between strategy and execution. When the intake form is vague or incomplete, every downstream team fills the gaps with assumptions, and those assumptions cost you rework, delays, and campaigns that miss what was intended. That form does more work than any other document in your production system.
Wyatt built a request form that captures up to 25 fields: target audience persona, copy requirements, naming conventions, tags, channels, and the campaign’s strategic objective. The form forces specificity before anyone builds anything. In Episode 80, he described the whiteboarding session that follows as the one genuinely human part of the process, where a strategist turns the brief into a concrete action plan that can then run largely automated through to delivery and reporting. Everything the automation does downstream depends on what that brief captured.
Elena Hassan, Global Head of Integrated Marketing at Visa Direct, described a different intake challenge in Episode 169: at enterprise scale, the creative brief doubles as a stakeholder alignment tool. Before her team builds anything for a global brand awareness campaign, she brings in partners from all 5 regions while the brief is still being written. They check whether the personas hold up and whether the messaging travels across Latin America and Asia or leans too hard on one market’s assumptions. Every region gets a voice at the start, and production runs faster downstream because the disagreements surface before the creative exists.
“It’s like being a chef in the kitchen with a lot of cooks, but you still manage to cook something that’s good.”
Elena Hassan, Episode 169
Jim Williams, who has led marketing operations at senior levels across multiple organizations, described the same problem one layer up, at planning, in Episode 146. Skip structured intake at the planning stage, where budget gets allocated into departments and campaigns get named, and you get what he calls “plan drift”: the original strategy erodes as individual teams make local decisions. The rally cry gets fuzzy, teams diverge, and by mid-year nobody is quite sure which campaigns are serving the annual objectives. The intake form is what connects individual campaign requests back to the plan.
The survey data supports spending more time on the brief. 36% of enterprise teams name coordinating across multiple teams and stakeholders as a top reason campaigns miss their launch date, and 69% cycle through 2 to 3 rounds of revision before anything gets approved. Revision rounds are where an underspecified brief shows up on the calendar. Every question the form failed to ask becomes a comment thread later, and comment threads are measured in days.
- Audit your current intake form against the last 10 campaigns. How many fields were left blank or filled with “TBD”? Those are the fields to make required, or to replace with a sharper question.
- Add a field that links every campaign request to a named strategic objective or annual priority. If a requester can’t fill it in, the campaign probably shouldn’t be built yet.
- Run your intake form through a new team member before the next launch cycle. If they need help answering questions, the form is too ambiguous for the team that has to execute against it.
Why naming conventions are the backbone of every campaign you measure
Campaign naming is one of the least glamorous parts of marketing operations and one of the most consequential. When naming is inconsistent, you can’t compare performance across regions, channels, or time periods. When naming is disciplined, reporting becomes something you can trust, and trust is what earns marketing ops a seat in strategic planning conversations.
Jim surfaced a specific failure mode in global marketing organizations: different regions invent their own vocabulary for the same concepts. One region calls the thing a “flight,” another calls it a “campaign.” Without a consistent hierarchy and shared language (programs, campaigns, tactics, activations), leadership can’t evaluate or benchmark efforts across the organization. In Episode 146, Jim argued that marketing ops is the right team to own this standardization, and that owning it is strategic work: when terms mean the same thing everywhere, leadership can make precise resource allocation decisions instead of guessing at what the data represents.
Ana Mourão, who led martech modernization at Stanley Black & Decker, described the operational consequence in Episode 159. When a major product campaign launches across multiple regions without standardized tracking nomenclature, comparing U.S. and Canadian performance requires manual intervention at every reporting cycle. The same data, named differently, creates a structural barrier to the cross-market analysis that actually improves campaigns. Her fix was a data template that standardized what fields got captured, how they got named, and who was responsible for enforcing consistency before a campaign went live.
“The success of any ABM or any demand gen tactic starts and stops with marketing operations. If you don’t have that excellence in how you bring things together, everything else is just secondary.”
Nadia Davis, Episode 184
Nadia Davis, who has led marketing operations and attribution strategy at multiple organizations, extended this argument in Episode 184: funnel stages, channel definitions, campaign naming rules, and shared language across sales and marketing are the operating system behind go-to-market execution. If those are vague, every attribution model built on top of them is unreliable before it even runs. Naming decisions are measurement decisions.
Knak’s survey shows how far most teams are from that. 69% measure email and landing page performance by click-through rate and only 41% by revenue or pipeline influenced, which makes marketing decision-makers 68% more likely to report on engagement than on business impact. That gap holds across industries, business models, and AI maturity levels. Roughly a third of teams consistently hit or exceed their performance targets. Email returns at least 36:1 for a third of companies, so the revenue story is sitting there for most teams to tell. What’s usually missing is the naming discipline that would let them assemble it.
| Naming tier | What it covers | Why it matters |
|---|---|---|
| Program | Top-level strategic initiative (e.g., Q3 Pipeline Push) | Connects budget allocation to outcomes |
| Campaign | A specific coordinated effort within a program | Unit of measurement for performance comparison |
| Tactic / Activation | Individual channel execution (email, paid, event) | Allows channel-level attribution without losing program context |
| Tracking parameters | UTM tags, platform IDs, merge field conventions | Links execution data back to the campaign tier for reporting |
How to build self-service campaign templates that scale without creating chaos
Self-service templates solve the same problem from 2 directions. For the marketing operations team, they eliminate the repetitive build requests that consume hours every week. For field marketers and regional teams, they eliminate the dependency on MOps for every campaign launch. Done badly, self-service creates a sprawl of off-brand, unmeasured campaigns. Done well, it multiplies what your team can ship without multiplying headcount.
Angela Rueda, Meta’s Director of Martech, described how her team rebuilt Meta’s content development in Episode 161. Before the change, asset creation ran through Google Sheets and endless email threads. After expanding Knak into their central content hub, hundreds of global marketers could build emails independently without ever opening the marketing automation platform. The template library became the production environment. Angela’s team set the guardrails, meaning brand standards, approved modules, and compliance requirements, then stepped back from individual build requests.
Ashleigh Johnson, a marketing technologist who built self-service systems at TrendMicro and Cornerstone, described the field marketing problem from the other side in Episode 132. Field marketers operate in a fast-paced, results-driven environment that frequently clashes with the structured pace of marketing operations.
Before she built the self-service process at TrendMicro, every field request required her to go into the system and build from scratch. After it, field marketers could launch independently from pre-approved templates, and the quality constraint shifted from Ashleigh’s availability to the template itself. Once the template carries the standard, your calendar stops being the bottleneck.
“Knak allows users to create emails independently without needing to delve into their automation platform. Users work within Knak, sync their work to their MAP, perform quality assurance, and then execute their campaigns.”
Vish Gupta, Episode 128
Vish Gupta, who built the campaign operations system at Databricks, described self-service as a question of sequence. In Episode 128, she walked through the Knak-based workflow as a deliberate separation of concerns: email creation happens in one environment (Knak), MAP execution happens in another, and QA is a defined step between them. Self-service only works if each step in that chain has a clear owner and a clear handoff. Give people the buffet, Vish said, but make sure the food is already prepared and you’re not handing over the kitchen.
Tool sprawl is what a good template library replaces. 54% of enterprise teams use 3 to 5 separate tools to produce a single marketing email, and when a new capability gap opens up, 38% patch it with whatever they already own. Knak’s survey also found teams are more likely to use dedicated tooling for landing page production than for email, even though email carries higher volume and more brand scrutiny. Every extra tool in the chain adds a handoff, and every handoff is a place where copy, creative, and tracking get out of sync.
The teams that move from idea to send in under 4 hours share a profile: 2 to 3 people involved in email production, 16 to 30 people on the marketing team, 3 to 5 tools in the stack, and AI used deliberately rather than experimentally. The results show up in production time. Amazon cut email production by 95%, from over 3 hours to under 10 minutes, by building a library of on-brand reusable modules in Knak and giving marketers direct access to the platform.
Andrew Eberting, Head of Global Martech and Operations at Amazon, framed the return in hours: “Time is the currency. If I can give my team hours back by removing friction, that’s the ROI.” Google Cloud reduced campaign change requests by 90% and booked up to $2M in savings. Uber rebuilt its global production model and cut asset deployment from 10 days to 4.5.
“Don’t read this as a shopping list. Yes, the data shows these teams tend to run Salesforce Marketing Cloud and ChatGPT, but those tools are only part of the picture. They’re also operating with small teams, fewer tools around the core platform, clear ownership, and AI aimed at the work that slows a launch down. Hand that same stack to a team that hasn’t done the operating work, and nothing magically speeds up.”
Sara McNamara, RevOps and Marketing Operations leader, in Marketing Production in the Age of AI
What these teams were building has since picked up a name. Knak’s CEO Pierce Ujjainwalla introduced the term “Marketing Production System” at Adobe Summit in 2026: the layer between AI tools and deployment platforms, purpose-built for the work that happens between a campaign idea and a live send. Knak’s April 2026 launch of their MCP Server made that layer programmable. AI assistants can now discover assets, search by brand or campaign, and generate emails inside a Knak instance, with the same rendering pipeline handling Outlook compatibility, dark mode, responsive behavior, and CSS that the visual editor would have handled manually. Automating the build gets you the same output through a different front door.
The same infrastructure gets wired very differently from company to company. OpenAI runs campaign requests through Slack, where a Codex agent structures the brief, creates a Linear ticket, generates the email in Knak, and routes it to a marketer for final refinement before deployment. Meta built an internal tool called Launchpad that feeds into Knak’s API, combining AI-generated content with brand controls before routing the asset through approval and into deployment.
Google positions Knak as the creation node at the center of every campaign, connected to an internal AI orchestrator and to downstream tools. Nvidia runs a design-to-deployment chain from Figma into Knak production, then through Viva Translate for AI localization before pushing to Marketo and Salesforce Marketing Cloud. 4 companies, 4 architectures, and all of them treat the production step as a node in a larger automated workflow rather than something a human rebuilds by hand every time.
Context access changes the output by a measurable amount. Knak’s research found that AI without access to brand guidelines, templates, and campaign structure produces drafts that are roughly 20% complete: technically coherent, but needing heavy rework on voice, structure, and brand alignment. With MCP feeding that context in, completion rates reach 80–90%. Same model, same prompt, and the only variable is whether the system could see your actual production environment. That gap is the argument for building your templates and brand governance before you build your AI workflows. The AI is only as production-ready as the context you give it.
The QA processes that catch problems before campaigns go live
Campaign QA is the part of production teams under-invest in until something goes wrong: a broken personalization token in a mass email, a webhook firing against the wrong segment, a campaign that passes every visual check and then renders as a jumble of boxes in Outlook. The instinct after a bad launch is to add more checklist items. The better fix is to change how the checklist works.
Justin Norris, a marketing operations consultant, described the QA failure mode in fast-moving environments in Episode 107. At a startup where he skipped thorough QA under time pressure, cleaning up meant building a new software application to reverse the incorrect data changes, a recovery effort that cost far more time than the QA would have. His baseline recommendation is peer review: someone who didn’t build the campaign checks it before it goes out. In agency settings, assigning QA to junior consultants pays twice. They learn the platform by reviewing real production work, and senior consultants get checked by someone who’ll ask “why does this work this way” instead of assuming it does.
Abby Gailey, who built campaign operations systems across B2B and B2C environments, introduced a framework in Episode 113 that flips the traditional checklist. A negative checklist is a list of specific errors and oversights that have already bitten you, things to actively hunt for rather than steps to confirm. Think of a pilot’s pre-flight checklist written as “do not skip the engine check” instead of “engine checked.” The value compounds: every post-mortem adds an item, and the list turns into institutional memory for how your campaigns break, drawn from your own failures rather than someone else’s template.
QA is also the stage where enterprise teams have automated the least. Knak’s survey found manual testing is still the default for email and landing page QA, which adds cost and calendar time that never appears on a project plan. April Mullen, Head of Content Marketing at Impel, described the alternative in the report: the strongest programs enforce required elements, accessibility checks, domain and link validation, and legal disclosures automatically, and reserve human review for higher-risk sends like regulatory communications or executive announcements. Automating the mechanical checks is what buys your reviewers the attention for the sends that actually carry risk.
“QA is like a meteorologist anticipating and identifying potential storms before they occur, ensuring the team can prepare and adapt to maintain smooth project progress.”
Abby Gailey, Episode 113
Teams using AI to generate campaign content need to add a step. 88% of teams in Knak’s survey say AI-generated marketing content needs moderate (57%) or substantial (31%) editing before it’s usable, and the reason is structural: 63% of them are prompting ChatGPT, which has no access to their brand guidelines, approved templates, or governance rules. Knak tested 3 leading LLMs (Claude Opus 4.6, GPT-5.4, and Gemini 3.1 Pro) on HTML email generation across 3 prompt complexities.
The combined average score across all 9 emails was 61 out of 100. No output scored above 71, and all 9 needed material rework before production use. Rendering was where it got ugly: 8 of the 9 AI-generated emails failed in Outlook for Windows, with incorrect widths, Times New Roman fallback fonts, broken dark mode, and inconsistent VML button support. 2 of 3 models triggered complete style block rejection in Yahoo and AOL because of CSS comments in the generated code. Every email that looked fine in a browser preview broke in multiple ways once it hit a real inbox. The browser-to-inbox gap that email marketers have managed for years survives AI writing the HTML, and it gets harder to catch, because the output looks plausible right up until it doesn’t.
Stripe’s version shows where QA automation is heading. As part of their Knak MCP implementation, they built a custom validation agent that confirms a final email matches the requirements of the original brief before it’s cleared for deployment. The agent checks intent rather than rendering or links: did the email that came out of the build process deliver what the brief asked for? That closes a loop most QA checklists never address, which is the drift between what was requested and what was built. A checklist confirms a campaign is technically correct. A brief-validation agent confirms it’s strategically correct. Most teams only have the first one.
Aboli Gangreddiwar, who builds AI agents for marketing operations, named a QA gap most teams have never addressed in Episode 191: journey QA. Most marketing automation platforms have some form of email QA tooling. Almost none have anything for validating the branching logic of a complex multi-step journey, whether the right contacts enter the right branches, whether the timing logic holds across edge cases, whether a suppression rule is firing at all. That logic only exists inside the platform, and unless someone builds a dedicated QA workflow for it, it gets tested against real contacts in production. The teams catching this first are prototyping AI agents that can trace journey logic end to end before any send happens.
- Run your QA checklist through a retrospective against the last 5 campaign failures. How many of them would the checklist have caught? If the answer is “not many,” the checklist is measuring compliance rather than risk.
- Convert your checklist from a to-do format to a negative format. For every item, rewrite it as a specific error to look for rather than a step to confirm.
- Build a separate QA step for journey logic, not just email rendering. Document the expected behavior of each branch and suppress conditions, then have a second person walk through it before the journey goes live.
How approval workflows in complex organizations slow campaigns down, and what to do about it
Approval workflows are where campaign production goes to die in large organizations. Legal review, brand review, compliance sign-off, regional stakeholder approval: each one adds calendar days to a production cycle that was supposed to take a week. The approvals themselves exist for real reasons. What’s out of date is the process around them, built for a world where campaigns were slower, fewer, and more consequential to get wrong. That world no longer exists for most teams.
Knak’s survey put a number on the cost. Getting approvals and sign-off is the single most cited reason enterprise campaigns miss their launch date, named by 47% of teams, ahead of design and creative production at 38% and cross-team coordination at 36%. It’s also the most addressable of the 3, because most of the elapsed time goes to routing and waiting rather than to the review itself. Reading a campaign takes a reviewer minutes. Getting it in front of them, and getting an answer back, is what takes days.
Elena described the scale of this problem after Visa acquired her previous company in Episode 169. Campaigns that used to roll out in days now take months, because the stakes are high and the process is rigorous: legal reviews, vendor contracts, compliance checks, regional approvals. Each layer was added to prevent a real category of risk, and none of them were built for speed. The operational lesson is to scope approval workflows to the risk profile of each campaign, so that audience size, channel, and content type determine how much review a send actually gets.
Aboli described the compliance bottleneck as the next frontier for AI agents in Episode 191. In regulated industries, legal teams comb through every campaign for risky language, improper data use, or brand missteps, and that process can stall launches for days. A compliance agent with the right guidelines codified can do a first-pass review in minutes, flagging likely issues for a human and clearing low-risk content automatically. Humans stay in compliance, and their judgment gets reserved for the actual gray areas instead of burned on campaigns that have always been fine.
“There is also the compliance problem. In many organizations, compliance reviews are the true bottleneck, especially in regulated industries. An agentic compliance layer can stall launches for days; an AI agent with the right guidelines can do that first-pass review in minutes.”
Aboli Gangreddiwar, Episode 191
Rich Waldron, CEO of Tray.ai, described the structural fix in Episode 162: complete approval workflows that move without human bottlenecks. The key word is “complete,” meaning a workflow that routes approval requests, tracks status, escalates when reviewers go silent, and moves to the next step automatically once approval lands. Most organizations have the routing and stop there, which is why approvals sit in inboxes until someone chases them. Teams recover the most calendar time by automating how the workflow moves, while the approval decision itself stays with a human. Knak’s survey backs the mechanism: advanced AI adopters are 56% more likely to use native MAP approval workflows than everyone else, which is a small structural choice with a large effect on the calendar.
Tiering by risk level cuts approval volume without cutting governance. Low-risk content like subject line variants and alt text gets AI generation with spot-checking. Medium-risk content like campaign copy and email bodies gets an AI draft that a human edits and approves. High-risk content like brand messaging, legal language, and regulatory disclosures stays human-owned, with AI in a research-assist role only. Over 70% of marketers have run into AI-generated content with hallucinations or off-brand material, which happens when teams point AI at high-risk unbounded tasks before they have the governance to catch failures. Tiering is what makes approval automation safe rather than reckless.
| Risk level | Content type | AI role | Human role |
|---|---|---|---|
| Low | Subject lines, alt text, meta descriptions | Generates | Spot-checks |
| Medium | Campaign copy, email bodies, social posts | Drafts | Edits and approves |
| High | Brand messaging, legal language, compliance disclosures | Research assist only | Owns creation |
The content supply chain: automating end-to-end campaign production
The most advanced version of campaign operations is a structurally different process, one where automation handles everything between the strategy decision and the performance report, and humans stay in the work that requires judgment. Speed is a byproduct of that design. The framing is no longer speculative, and the teams building toward it are already seeing pieces of it work.
Wyatt’s team was building exactly this in 2023. In Episode 80, he described a system where a campaign brief comes in through a structured request form, goes through a whiteboarding session with a strategist, and then flows automatically through to delivery and reporting without a developer or marketing operations person stepping in. He called it a “content supply chain,” briefs moving through marketing automation all the way to delivery, and his team had been operating it for 6 months. The automation handles the batch work. The humans own the strategy and the whiteboard.
Alison Albeck Lindland, a marketing executive who has worked across AI-driven personalization platforms, described the content supply chain from the production output side in Episode 181. She described it as replacing “the outdated, print-era process of campaign production,” where teams manually assembled and formatted content for every send, with a system that builds and refreshes content on the fly. AI-powered content decisioning generates one-to-one experiences for emails and mobile messages automatically, selecting from a catalog of approved content modules based on customer context. The marketer’s job shifts from “assemble this campaign” to “define the rules and the catalog.” The system absorbs the production work.
“AI powered content decisioning lets you replace outdated campaign production with a system that builds and updates personalized content automatically. That way you can stop wasting hours on manual assembly, focus your team on higher-value strategy, and deliver emails and mobile messages that improve with every send.”
Alison Albeck Lindland, Episode 181
Jeff Lee, who built the martech infrastructure powering Calm’s billion-message machine, grounded what that scale actually requires in Episode 158. Marketing automation platforms operating at this level are engineering systems, combining complex backend infrastructure, real-time data processing, and experimental feedback loops. Teams building toward a true content supply chain need to stop treating their MAP as a campaign builder and start treating it as a product. The infrastructure decisions you make for reliability, data freshness, and scale set the ceiling on what the production system can do.
Vercel is running a version of this today. Cory Gabor, their Senior Marketing Operations Manager, makes the point in Knak’s report that the shift required no new tools at all. Slack, Knak, Linear, Customer.io, and Salesforce were already in the stack. What changed is that an agent now carries the work between them.
“Our campaign stack, Slack, Knak, Linear, Customer.io, Salesforce, used to be a set of destinations a human walked between, carrying data by hand. Now it is an assembly line that runs from a single Slack thread. The tools did not change. What changed is that an agent now connects them and carries the work between them, so a person does not have to.”
Cory Gabor, Senior Marketing Operations Manager at Vercel, in Marketing Production in the Age of AI
The ROI case for this kind of infrastructure is no longer anecdotal. Forbes built AI assistance into their content production workflows and saved 18,000 hours annually while doubling landing page conversion rates, with the AI embedded in the production platforms rather than bolted on as a separate step. Block, the financial services company behind Square and Cash App, connected MCP servers across Snowflake, GitHub, Jira, Slack, and Google Drive and reported employees saving 50–75% of time on common tasks, with work that used to take days finishing in hours. In both cases the gains came from AI operating inside the production system with access to real context, which is a different thing from prompting a tool in isolation and integrating the output by hand.
GrowthLoop pushed the architecture forward in 2026 with their composable AI decisioning launch. They use causal AI where most AI-assisted campaign tools use pattern-matching: the system builds a causal context layer over your warehouse or data lake, and every campaign you run makes the next decision smarter. It’s also zero-copy, so data stays in your warehouse instead of moving out to fuel the decisioning. That keeps the data layer coherent rather than fragmented across execution environments. The direction of travel is away from campaign-based thinking, with its discrete sends, batch workflows, and one-off builds, and toward continuous real-time decisions that adapt as signals come in. The production system stops being a pipeline from brief to send and becomes a loop.
There’s a timing argument for building this now. 88% of marketing teams produce campaign assets for email, making it the most widely used channel by a wide margin, and 52% plan to add or substantially increase email spend over the next 12 months. Scaling volume through a process that already takes a full business day and 4 people per send multiplies the friction along with the output. Knak’s conclusion is that what teams need is a dedicated production layer that sits between the campaign idea and the MAP, because the MAP was built to deliver rather than to build. Get that layer in place before you scale the channel through it.
How AI is shifting campaign operations from execution to orchestration
The campaign operations job is changing fast. The tactical execution work, building lists, setting up sends, configuring suppression logic, checking renders, is the most automatable part of the role. The teams already using AI to absorb that work are redeploying their campaign ops people toward what automation can’t do: reading signals, adjusting strategy, and maintaining the systems the automation runs on.
Knak’s survey shows where teams actually sit today. 70% have deployed AI in some capacity, with 29% reporting advanced adoption where AI is embedded across the entire workflow, 41% intermediate, and 25% still early. Teams at $1B+ organizations are 72% more likely to be in the early stages, because scale brings governance requirements and legacy infrastructure that slow adoption down. The most common production uses are generating first drafts of email and landing page copy at 64%, generating or editing images at 56%, analyzing performance data and suggesting optimizations at 56%, writing subject line variants at 48%, and brand and compliance checks at 41%. Most of that sits at the drafting stage, which is where the returns are smallest.
The gap between advanced adopters and everyone else is already measurable, and it tracks culture more than budget or headcount. Advanced adopters are 62% more likely to produce a single marketing email in under an hour, 25% more likely to send 100 or more emails a month, and twice as likely to be very satisfied with their email and landing page performance. They’re also 44% more likely to use AI agents for building and coding emails and landing pages, 41% more likely to use it for brand and compliance checks, and 34% more likely to use it for translation and localization. That’s AI operating inside the production layer, past the first draft. When a capability gap opens up, 56% of them buy a dedicated tool, against 36% of all respondents. They’ve worked out where AI reaches its limit and purpose-built tooling takes over.
Chris O’Neill, CEO of GrowthLoop, described the shift in Episode 177 as the difference between “waterfall marketing” and agentic marketing. Waterfall marketing: a marketer decides on a goal, builds a segment, builds a campaign, sends it, analyzes results, starts over. Agentic marketing: a swarm of AI agents runs segmentation, content variation, send time optimization, and feedback collection in parallel, adapting in real time. The agents execute the steps, and the marketer sets the objectives, defines the guardrails, and owns the outcomes. Chris was explicit that legal and compliance teams can stay fully involved without killing velocity, because compliance logic gets codified into the agent constraints instead of reviewed at every campaign launch.
Pini Yakuel, CEO of Optimove, described the speed implication in Episode 81: the first job for a marketer in a self-optimizing environment is making the trip from idea to execution short. Marketers, Pini observed, are an impatient bunch. They want to go from idea to live test in minutes or hours rather than days. When the production system makes a marketer build, configure, QA, and launch every experiment by hand, the bottleneck sits in the pipeline between the idea and the data rather than in anyone’s creativity.
Zapier launched AutomationBench in 2026, a public benchmark that measures how well AI models perform real business tasks in real tools instead of abstract reasoning puzzles. Early results showed no model clearing even a 10% success rate on multi-step business workflows like updating CRM records or sending accurate follow-ups without errors.
The takeaway is to build carefully and scope tightly. Some campaign operations tasks are bounded: repeatable, measurable, and light on institutional context. Subject line generation, alt text, merge tag formatting, and render checking all fall here, and AI handles them reliably. Other tasks are unbounded, requiring institutional judgment that shifts with audience, competitive context, or regulatory factors. Campaign strategy, brand voice decisions, and compliance review sit in that group, where AI assists and a human owns the call. The teams getting the most out of agentic campaign operations are precise about which category each task falls into, because treating an unbounded task as bounded is how you get generic output, brand drift, and compliance failures at scale.
BCG’s analysis of MCP adoption makes the infrastructure argument in a single line: without a standardized integration layer, integration complexity in a marketing stack rises quadratically as you add tools. With one, complexity rises linearly. That’s the structural reason the best campaign production systems are investing in MCP connectivity over one-off integrations. It’s the only architecture that stays manageable as the stack grows. It also changes how you should think about AI agents running inside your production system. Knak’s guidance is to treat them like junior employees with elevated access: clearer scope definitions, stricter permissions, full audit logs, and no irreversible actions without approval. The governance you build around your agents matters as much as the agents.
The access layer for campaign production data is changing too. MoEngage’s MCP server launch shows where the category is moving: AI tools can now pull campaign performance data, audit configurations, and check email, push, and SMS details in plain language. “Which campaigns had the highest CTR last quarter” has become a prompt rather than a dashboard build. That matters for production operations because it lowers the cost of monitoring and auditing what the system is doing. When a campaign engineer can interrogate their production pipeline through conversation instead of reports, the loop between what shipped and what to fix gets much shorter.
“To keep your campaign operations job as AI continues to knock, immediately shift your focus from tactical execution to strategic functions. Master business alignment skills, develop creative decision-making capabilities, and build continuous optimization programs.”
Episode 168: AI’s talent crunch
The talent implications, explored in Episode 168, suggest a specific career strategy for campaign operations practitioners. One-off campaign execution work, meaning repetitive builds, batch-and-blast sends, and manual list pulls, is “at risk” in AI terms. The AI-resistant position is campaign ops built around continuous optimization: always-on programs that need constant monitoring, adjustment, and improvement rather than a recurring cycle of build-and-send. The shift is from campaign producer to campaign architect. The production system runs the campaigns. You design and improve the system.
Practitioners are already pricing this in. 53% of the marketing decision-makers in Knak’s survey named driving AI and automation adoption across the organization as the most critical skill for their career growth over the next 1 to 2 years, more than 3 times any other capability on the list. At $1B+ companies it’s 64%, and at organizations with advanced AI adoption it’s 72%. The people closest to production systems are the ones betting hardest on this skill, which is the clearest signal in the report about where the job is going.
Paul Wilson, a marketing operations strategist, named this role in Episode 104: the campaign engineer. Campaign engineers bridge marketing strategy and operational execution. They understand both the strategic intent behind a campaign and the technical constraints of the system that will run it, and they can sit in the whiteboard session and then design the automated workflow that turns it into a live campaign without losing anything in translation. As production automation absorbs the tactical execution work, the campaign engineer gets more valuable. The bridge between strategy and systems still needs someone who understands both sides.
Survey data throughout this playbook comes from Marketing Production in the Age of AI, Knak’s 2026 report. The research was a 30-question survey of 333 marketing decision-makers at organizations with $50M to $1B+ in annual revenue across the US, UK, and Canada, fielded between April 29 and May 18, 2026, and conducted by Datalily via Centiment on behalf of Knak.
Want more practitioner perspectives? Subscribe to Humans of Martech for weekly conversations with the operators building the future of marketing technology.