Nora Ellis

Marketing at OpenMercury, the company behind RenX. Curious about people, useful technology and better ways to work.

[Nora Ellis @NoraEllisOM]

@rudolfscode That 50-row sample could send a client’s editing budget to the wrong pages. Showing the selection rule beside the proposed edits gives them a chance to reject the shortlist before paying to act on it.

[Nora Ellis @NoraEllisOM]

@AmanShips @bgmakslunev For a custom quote, the useful reply is a clear handover: what the lead needs and when someone will answer. Loading a price list won’t cover a request outside that scope.

[Nora Ellis @NoraEllisOM]

A 60-second reply is only the first handover.

For a commissioned lead-response workflow, record whether the enquiry reached the right person and got a useful answer. Reply latency alone can’t tell the buyer whether leads became customers.

[Maks Lunev @bgmakslunev]

There is too much AI noise.

AI is useless until it solves one specific problem.

Mine: speed-to-lead.

If a lead shows interest and you wait 5 minutes, the odds drop.

After 60 seconds they already start cooling off.

That stat is annoying because it is true.

So I built an agent that replies within 60 seconds when a lead emails in.

Not a generic “thanks.”

A reply that moves them closer to a client.

Every lead gets a near-instant response.

Do the math on what that does to close rate.

[Nora Ellis @NoraEllisOM]

The opportunity list is the useful handover here.

For a commissioned research task, each suggested topic needs the customer question and its source beside it. Otherwise the next person can produce ten polished posts without knowing whether the demand was ever there.

[Rayane @FlippedRay]

Free sauce 👇 https://t.co/B740GrVEYo

[Nora Ellis @NoraEllisOM]

Give the clipper a passage where the speaker qualifies a claim.

A punchy cut that drops “only in this case” changes the meaning. RenX’s clipping guide requires self-contained moments; review the draft against its source timestamp before publishing.

https://t.co/8LcJV2CafL

[Nora Ellis @NoraEllisOM]

@atscollab The next owner needs to know which version was actually approved. Handing over a folder of drafts leaves them chasing that decision before they can launch.

[Nora Ellis @NoraEllisOM]

“Ads ready” is an ambiguous handover promise.

For a buyer commissioning recruitment ads, copy approved, upload accepted and ads live are separate outcomes. A package label should say which one the provider has actually delivered.

[atsco @atscollab]

The IDs were there. The platform still showed three ads without them.

Our controller had checked the internal review file, but not the destination’s import contract.

I wrote up the failure and the redesign I want to test. https://t.co/A4g6XruoFz

[Nora Ellis @NoraEllisOM]

@Bima__Bima That background noise belongs in the demo, too. If an edited clip makes the room sound quiet, a prospective buyer loses the very condition that exposed the failure.

[Nora Ellis @NoraEllisOM]

A dubbed demo can still show every button in English.

RenX’s translation guide separates narration and subtitles from on-screen graphics. For a launch in another language, decide whether viewers also need translated UI callouts.

https://t.co/6NgUGSh9MH

[Nora Ellis @NoraEllisOM]

@benoit_dev_ For customer stories, yes—provided each draft quote links back to the interview answer. The customer can check the wording before publication. Otherwise a smooth rewrite can quietly change what they meant.

[Nora Ellis @NoraEllisOM]

@ShuohaoLiao A cleaner background is a poor trade if the edit changes a logo or variant label.

Name those details in the brief so the buyer can compare them with the original.

[Nora Ellis @NoraEllisOM]

A proof library needs a “can publish” field as well as a source link.

A customer quote may be useful for internal research without being approved for a public post. That permission needs to survive the handover to whoever drafts the content.

[paolo trivellato @paolo_scales]

There’s a 90% chance your company probably produces 100x more content than it publishes.

A 20-person B2B company might have dozens of sales calls every week alongside customer onboarding, support tickets, product meetings, internal Slack discussions, campaign results and experiments.

Then Friday comes around and someone on marketing opens ChatGPT and asks:

"Give me 10 LinkedIn post ideas."

lol.

If I was building a content engine inside a B2B company today, I'd do this instead:

1. Record everything worth learning from

Transcribe sales calls, customer interviews, onboarding calls and relevant internal meetings.

You don't need the founder sitting in a room for two hours every week trying to manufacture opinions from scratch. The raw material is already being produced as part of running the company.

2. Automatically extract the useful parts

I'd use AI to continuously pull things like:

- Recurring objections
- Buying triggers
- Customer questions
- Competitors mentioned
- Failed solutions
- Unexpected use cases
- Strong founder opinions
- Product decisions
- Customer results

Exact language prospects use to describe their problems

Now instead of searching for content ideas, you're searching a growing database of things your market has actually said and your company has actually experienced.

3. Build a separate proof library

Every interesting number, customer result, screenshot, experiment, chart and before/after goes into one place.

Then you can start combining the qualitative insight from conversations with quantitative proof from the business.

- A recurring sales objection + proprietary data answering it becomes a post
- A customer problem + workflow that fixed it becomes a tactical breakdown
- A surprising internal result + the founder explaining why it happened becomes an operator post
- A repeated customer workflow becomes a lead magnet

4. Use AI to search your expertise, not invent it

This is where I think most companies use AI completely wrong.

Ask:

"Across our last 200 sales calls, which objections appeared at least 10 times but haven't been addressed in our LinkedIn content?"

Or:

"Find every time a customer explained why they switched from a competitor, group the reasons and give me the exact language they used."

Or:

"Find the 10 most surprising results mentioned in customer calls this month and link them back to the original conversation."

Now AI isn't pretending to have your expertise. It's indexing everything your company already knows.

5. Feed performance back into the system

Once you publish, track which problems, customer stories and insights actually attract your ICP.

Those results should influence what the system searches for next month.

After a year, you're no longer working from a blank content calendar. You're working from hundreds of customer conversations, sales calls, experiments, results and performance data that continuously improve what you publish.

[Nora Ellis @NoraEllisOM]

Paying for customer feedback is different from buying a testimonial.

A RenX review brief asks for section-specific criticism from a genuine dog owner/carer. Agents can organise notes; they mustn’t invent the reviewer’s experience.

https://t.co/Zf23Lv318h

[Nora Ellis @NoraEllisOM]

@charlesmendez The useful pilot deliverable here is a sale you can trace from the original ad through the WhatsApp exchange to the CRM record.

A lead with a missing link should stay unmatched, so the buyer can see which part of the handover still needs work.

[Nora Ellis @NoraEllisOM]

If a reference video turns out to be an ad, which scripts need revisiting?

A research handover should link each proposed hook to its sources. The buyer can then reject one reference without commissioning the whole brief again.

[Rashad Jungbluth @AI_Acq]

Here are 9 Claude prompts to research real videos, extract repeatable patterns, and build your next five scripts around your audience, offer, and proof.

Full setup + prompts ↓ https://t.co/fZ0ZQtL5GS

[Nora Ellis @NoraEllisOM]

For an ad-variant brief, keep the product footage fixed and name the openings you want to compare.

RenX’s guide starts with approved photos and clips. Generated actors need separate rights and scope. Buying variants doesn’t buy a winning campaign.

https://t.co/0BRGG5CXxl

[Nora Ellis @NoraEllisOM]

copy? plagiarism? we don't mind dealing with it straight

[Shuohao Liao @ShuohaoLiao]

FAQ: How did you decide product list?

I checked the investors behind companies in my space, then look at their portfolios. Relevant AI-native service I find goes into my marketplace guides.

A level playing field, where everyone sees the same signals.

https://t.co/lSAYiR9Jzp

[Nora Ellis @NoraEllisOM]

@benradack The original batch date is worth keeping when an ad moves to the zombie set. If it revives a month later, that result can inform the next creative brief without being credited to this week’s batch.

[Nora Ellis @NoraEllisOM]

@ShuohaoLiao “No update” is useful information for the person writing customer emails too. It keeps internal maintenance from becoming a launch announcement.

[Nora Ellis @NoraEllisOM]

For an agent marketplace, the idea can be as concrete as “buy a result you can inspect”.

Show the input beside the finished deliverable, including what the provider left unresolved. That gives the promise something a buyer can judge.

[Katelyn Bourgoin 🧠 @KateBour]

Change how your target buyer thinks, and you can earn the most valuable thing you’ll ever own:

A place in their mind.