Your AI writes faster. Your emails still go out late.

A new report from Knak, the campaign production platform, asked marketing teams where their time goes. The answers should worry any business owner who has paid for an AI subscription and expected to get hours back.

85% of teams surveyed missed at least one campaign launch date in the past year because of workflow problems. 82% still spend at least half their time on production rather than planning. Kath Pay, CEO of Holistic Email Marketing, wrote the findings up for MarTech in September, and her conclusion was blunt: AI gives marketers more marketing, not more time.

I agree with the diagnosis. But most coverage stops at "rethink your workflow," which isn't much use to a founder on a Tuesday morning. So this piece covers what to change, in what order, and where AI belongs once you've changed it.

What the research found

64% of teams now use AI to write first drafts of emails and landing pages. So drafting got faster. Nothing else did.

The biggest delay, named by 47% of respondents, is getting approvals and sign-offs. Design and creative production came second at 38%, and coordination between teams third at 36%. 60% of teams need four or more people to produce one email, and 69% go through two or three rounds of revisions before anyone approves it.

And the drafts themselves still need work. 70% of teams have AI in their production process, yet 88% say its output needs moderate or substantial human editing.

Put those numbers together and the picture is simple. Teams used AI to reach the approval queue faster, then sat in it.

Read the source before you quote it

Knak sells software that manages email and landing page production. A Knak report finding that production workflow is the bottleneck suits Knak's sales team very well.

That doesn't make the numbers wrong. They match what I saw across years of running client services in agencies, where the brief took an hour and the sign-off took a fortnight. But look at who Knak asked: more than 300 enterprise marketing leaders, at a company whose customers include Google, Amazon and Uber.

One finding needs more caution than the rest. The report says teams that finish an email in about four hours involve fewer people and use AI deliberately. But smaller teams may run simpler campaigns with less riding on them, so treat it as a correlation and don't promise anyone a four-hour email.

The SME version of this problem

Knak surveyed enterprise teams, where four or more people touch every email and each approver adds their own round of comments. Large companies have that problem. Most SMEs don't.

In a business of 20 or 50 people, the approval chain is often one founder. They review every email, usually late, and rewrite the copy themselves because it doesn't sound like them. Nothing goes out until they've looked at it, and they've got a business to run.

The founder bottleneck is harder to fix than a committee, because the person causing the delay is the person paying for the fix. And no AI tool will change it. The founder has to decide what they no longer need to see.

Fix one: approve the brief, then leave the draft alone

Most revision rounds start as copy edits and turn out to be arguments about strategy. The reviewer reads a draft, realises they wanted a different message or audience, and sends it back with "can we make it punchier?"

So move the argument earlier. Before anyone writes or prompts anything, agree a one-page brief: who the email is for, the one thing they should do, the offer, and how you'll measure success. The founder signs that off.

After that, reviewers can reject a draft for breaking the brief. They can't reject it because they've changed their mind. In my experience this single rule removes a full revision round, though I'd measure it with your own campaigns before claiming a number.

AI helps here too, and few teams use it this way. Before drafting, paste the brief into Claude or ChatGPT and ask it to challenge you: who exactly is this for, what do they believe now, why would they act today? Ten minutes of hard questions at the brief stage saves the rounds Kath Pay describes, where a team tries to bolt the strategy on after the copy exists.

Fix two: tier your approvals by risk

A monthly newsletter built from an approved template carries almost no risk. A pricing change or a claim about what your product does carries a lot. Most businesses review both the same way.

Set up two or three tiers instead. Routine emails using approved templates and approved claims go out after a brand check, with no founder sign-off. Standard campaigns get one named approver with a 24-hour window, and silence counts as approval.

Only new offers, pricing, legal wording or anything unfamiliar goes to the founder. The founder sets the rules once and then stops reviewing tier-one work. That's the behaviour change, and it'll feel uncomfortable for the first month.

Where AI belongs in the approval process

The Knak report shows AI sitting at the start of the process. The better place for it is just before human review.

Load your brand guidelines, tone rules, banned phrases, approved product claims and any legal wording into a Claude Project, a Custom GPT or a Gemini Gem. Run every draft through it before a person sees it. It'll catch the wrong product name, the off-brand opening line and the missing terms, which are the nitpicks that burn whole revision rounds.

Be clear about the limits. These tools miss things and sometimes flag problems that aren't there. A person still owns the decision to send, but that person now reviews a cleaner draft.

Stop generating twenty options

AI makes options cheap. Kath Pay's point is that choosing between them still costs time, and more options can make it worse. Line up ten subject lines and your team will argue over which one sounds friendlier instead of asking whether the email should go out at all.

Cap what reaches the approver at two or three. Let AI generate as many as you like, then have one person filter them against criteria written into the brief. If the brief says the subject line must state the discount and the deadline, most options fail before anyone debates them.

Testing instead of debating sounds sensible, but SME email lists are often too small for it. By my own calculation, spotting a lift from a 20% to a 21% click or open rate with standard statistical confidence needs around 25,000 recipients per version. So test ideas that differ a lot, and judge them on clicks.

Opens won't tell you much about subject lines anyway. Apple's Mail Privacy Protection, introduced in 2021, inflates open rates.

Cut the editing by fixing what goes in

If 88% of teams still edit AI drafts heavily, the drafts probably started from thin prompts. "Write an email about our autumn sale" gets you generic copy, which someone then rewrites from scratch.

Give the model proper context. Write a voice guide with real examples of how you sound, a list of words you never use, your approved product claims, and your five best-performing emails from the past year. Save it once in your AI tool so every draft starts from it.

Then ask whoever edits the drafts to note what they changed for a month. Patterns show up fast: the AI opens too formally, buries the offer, overuses exclamation marks. Turn each pattern into a rule in the voice guide, and editing time should fall month on month, though I wouldn't expect it to reach zero.

Decide where the saved hours go

The report found that time saved by AI turns into more output: more campaigns, more versions, higher expectations. Most businesses never make that choice. The extra hours fill up with volume because nobody booked them for anything else.

Sometimes more output is the right call. If an extra segmented email makes money, send it. But make the decision on purpose.

Before you roll out a tool, say how many hours it should save and where those hours go, then put that time in calendars as strategy or customer research. Set a ceiling on send volume too. When unsubscribes and spam complaints start rising, you've reached it, whatever your tools can produce.

Measure the process for ten campaigns

Businesses track opens, clicks and revenue after an email goes out. Few track how long it took to get there.

Start a shared spreadsheet. For your next ten campaigns, log when each one enters and leaves every stage: brief, draft, review, design, build, test and scheduled. Track the number of revision rounds and the number of people who touched it.

I expect you'll find the work spends days waiting and minutes being written, but let your own data confirm it before you act. Once you know which stage holds work longest, you know where to put AI, and where a rule change will do more than any tool.

Where to start

Do the measurement first. Ten campaigns with timestamps will tell you more than any vendor report, including this one.

Then agree the brief template and the approval tiers. Both cost nothing and neither needs new software. Add the AI pre-flight check and the voice guide once the process holds, because AI placed on a broken process only gets you to the bottleneck sooner.