First Published: 17 August, 2026
Streaming in from the US, Britney Muller joined Digital Day Out as Founder of Orange Labs and one of the world's leading voices on AI and marketing. With a background that includes Marketing Manager at Hugging Face and Senior SEO Scientist at Moz, Britney has built a career translating complex AI and machine learning concepts into practical, usable strategy for marketers.
In her session #DDO2026, "The AI Execution Gap: Agents, Workflows and What's Next," she cut through the noise on AI driven search, sharing what actually moves the needle right now and handing over three live, steal able workflows you can put to work immediately. She left the crowd with plenty to think about, and plenty of questions too, more than she had time to answer on the day. So she's back to finish the job 👇
1. Are brand mentions that aren't linked still as successful as linked mentions?
Yes! There's an AI startup (in the film space) that took out a six-figure ad on a super niche industry site just to get a mention — because it was being cited so heavily in AI search results. No website link, no CTA, just wanted the mention, and lo and behold, it's started driving them business.
2. What's your advice for small teams trying to connect multiple AI tools into practical workflows without technical resources?
Be careful about what access you give AI tools. In development, there's "read" vs. "write" access: read means a system can only view information, no edits or changes; write means it can also edit or send changes. When connecting AI to email, for example, make sure you really trust the provider (Anthropic would be my recommendation) and start with read-only access until you're comfortable giving it more permissions, like drafting replies or sending follow-ups. Always keep a human in the loop.
3. How do you balance strong data governance with the need to move quickly on AI initiatives?
Ooof, such a great question! This is really hard, especially since most small-to-medium companies (agencies included) don't have much foundational data governance in place — data's usually scattered across servers, platforms, and systems. One thing I've seen work well is coming at it from the AI initiative side first: build an ugly proof-of-concept that feeds in your data, connects to APIs/external sources, and produces a desirable result. That gets buy-in to improve your data governance. That said, most people don't have access to all the data they'd need for a proof of concept — that's where dummy data comes in handy. Have an LLM recreate a sample of possible data to showcase your concept, then use that momentum to push for real data improvements.
4. Can we build most agentic AIs on the free versions of the LLM models?
Honestly, I'm not sure. You might be able to do that at some basic level, but to see true agentic AI behavior, the kind you can actually control and give lasting instructions to, you'll want at least the lowest paid tier of an LLM. Claude Cowork is genuinely mind-blowing for this, and I believe it's $20/month.
5. Are some countries strict on governance rules around AI and how it's used?
Yes, seeing more AI governance and guidelines emerging in the EU, though I'm not familiar with the specific details.
6. What's your favourite AI model so far that you feel is the most advanced for your work?
Fable, but it's costing me too much money, so I have to be really selective about when I use it. You don't need a blowtorch to light birthday candles, and you don't need Fable to write an email. I try to save it for when I'm building more advanced applications and tools.
7. Do you have one place to find open source AI resources to increase productivity?
I'm always checking labs.google (to see what Google's cooking up — that's where I found NotebookLM early on, which made me way more productive on several projects) and kaggle.com/competitions to see what bigger companies are trying to solve. I also love exploring the Hugging Face Model Hub for open-source solutions.
8. How much can be accomplished without paying for dozens of different AI tools? Are the free options good enough to start testing?
Honestly, no. Just pay for the entry-level Claude Cowork tier and start experimenting; your mind will be blown.
9. Are there any prompt websites you recommend (e.g. Prompt Cowboy or similar) and any tips and tricks?
Recommend avoiding prompt websites; they're a bit gimmicky. Your best prompts will come from you, because you have the specific context, data, and examples needed to solve your particular tasks. Here's a Prompt Engineering Course I did that's helped a lot of people. Pro tip: Use the LLM to help you generate the best, most optimal prompt (meta, right?!).
10. Understanding competitor brand health compared to ours in the digital space is a repetitive task — an agent that gives me a dashboard is something I've toyed with.
Ask yourself how the data was acquired and whether it's reliable. LLMs love to make things up and can drift over time if you're not monitoring them closely. Always double-check before using a dashboard to make big decisions or share it with stakeholders.
11. How do you balance using AI to turn customer insights into actionable tasks while protecting customer privacy (working in the health sector)?
Anonymize the data; remove PII, and you can even alter other patient-specific numbers or details. Also, always have people review large datasets like this line by line; dumping a huge dataset into an LLM and asking it to "do X" is a recipe for errors, obvious or otherwise. A better approach: pipe the data into a spreadsheet and have an LLM analyze each entry individually; one column for sentiment (positive/negative/neutral), another for specific topics — then analyze those results at scale.
12. If you were starting today, which tool would you say is your favorite AI, and why do you think other AIs haven't replaced it yet?
Claude Code is unreal. Try Claude Cowork first if you're non-technical (though non-technical folks can absolutely use Claude Code too! people just tend to get intimidated by the terminal. Don't be!). No other AI company is putting out releases anywhere near as valuable as Anthropic's right now, in my opinion. They're crushing it.
13. Love the idea of using Claude for Cowork. I've been trying to use it for exactly that — any specific prompts that would help?
Check out the Prompt Engineering Workshop I did. Pro tip: talk to Claude in natural language and let it help you build your perfect prompt (meta, I know). There’s also been hundreds of prompt studies done and the biggest finding to improve AI outputs is “few shot learning” (fancy AI terminology for “feed the AI a few examples” )
14. From your experience, what works best for separating fact from fiction? Seems like AI literacy and planning are essential to get true ROI.
The AI hype and misinformation could be AI's downfall and lead us into another AI winter. You're spot on, AI literacy is fundamental to using AI in qualified ways, versus unqualified ways that waste time, resources, and energy.
15. What do you see as the biggest reason AI initiatives fail after successful pilots?
No one maintains them. AI workflows and tools are programs; someone needs to understand them and check for drift, bugs, etc. Or, the initiative wasn't solving a truly important or impactful problem, which happens more often than you'd think. The best AI initiatives are often boring, simple applications.
16. Do you think Claude is far superior to ChatGPT? If we only had GPT, would we be missing out?
Yes.
17. If you were joining a large enterprise today, what's the first AI capability or operating model you'd build that most organisations are overlooking?
I'd start by figuring out what specifically moves the needle for the company: what tasks are necessary to make that happen, which are redundant or time-consuming, and which provide the most value. Prioritizing by value ensures an AI workflow drives real momentum. Counterintuitively, you have to start really small and build from there.
18. Definitely great to automate smaller tasks to make life easier!!!!!
Yes!!! It's such a thrill when you build a solution for a problem or task you actually have.
19. Have you observed any relationship between paid media investment and AI search visibility? E.g., if paid campaigns increase branded search and awareness, does that indirectly improve how often AI assistants recommend a brand?
Yes! There's an AI startup (in the film space) that took out a six-figure ad on a super niche industry site just to get a mention — because it was being cited so heavily in AI search results. No website link, no CTA, just wanted the mention, and lo and behold, it's started driving them business.
20. How do you best track how you appear in AI overviews and LLM answers as a brand?
I don’t worry about that & focus entirely on the channels currently driving me leads: LinkedIn and email are my primary distribution and lead-gen focus. Try to create high-quality shareable content so my audience can help me increase my brand mentions online.
21. What do we need to be considering from a data governance perspective? A lot of NZ brands don't have AI policies in place — what are the risks, and how can/should they protect themselves?
Start really small ; you don't need a big, perfect data governance policy to build an AI agent that supports your weekly newsletter. Once you build credibility and momentum with smaller AI solutions like that, you can start shaping a wider AI policy. Working with AI is highly iterative, and your policy will likely change over time & that's okay!
22. Love the workflows around internal linking and alt tags — tedious tasks made easy, with quick proofing after. Are these still worthy in AI search results? Do AI bots still trawl these things?
Both internal linking and alt tags improve your odds of ranking higher in traditional organic search, and traditional search largely powers AI search. So yes, I believe they're still worth implementing; especially now that it's easier than ever to identify and create them with AI's help. :)
23. Love your work, Britney! With marketing campaigns, what's your recommendation for measuring business outcomes and ROI on spend meant to enhance AI discoverability? The work may or may not drive changes in AI recommendations, and different people can get different brand recommendations from the same platform.
You're exactly right, it's probabilistic! Not a clean 1:1 with AI search. Peec.ai is doing some really savvy work in the AI discoverability space, but you could also build your own solution using Claude plus the SERP API and OpenAI API to see what AI outputs are surfacing for keywords or queries you care about (run each query 10+ times to get a rough sense of brand presence %). One caveat: we don't actually know what people are asking AI search, and measuring results (to your point) isn't straightforward, so hold any results from AI visibility tools (or systems you build yourself) very loosely.
24. What layer would you say is traditional SEO, and what falls under AI SEO?
SEO and AI are completely different technologies, quite literally different solar systems! AI doesn't index or rank websites; it doesn't even store the URLs of the pages it trained on. AI simply generates the average of what it's read on a topic. Search engines do the heavy lifting behind the scenes of providing AI with real-time information to improve AI outputs.
Do good SEO and digital PR, and you'll likely show up more in AI outputs.
25. You mentioned having the courage to action imperfect automation, and that resonates. But I find myself watching every step of an AI workflow instead of letting it run — I get nervous about it touching our CRM or customer data and making a costly mistake.
That's a good instinct! Stay cautious! Triple-check the system and remain skeptical. You can also anonymize data you're worried about and always keep a human in the loop.
26. How did you get comfortable making that leap? Was there a point you stopped needing 100% human oversight? Any practical advice for building enough trust in AI to let automations run while still managing risk?
Brilliant question! Honestly, I don't know... but the things I have AI doing for me aren't big or consequential tasks, I'm not a doctor (..can you imagine?). You do have to ask: what happens if AI gets this wrong? And if the answer is something really bad (like a patient getting the wrong treatment), don't use AI for that task. Always always check outputs before deploying results (human-in-the-loop, always). I've only had AI send or post things as experiments, and I listen to my gut. After automating my LinkedIn outreach, that experiment that got an 80% response rate, I felt gross. Which was a clear sign to me not to use AI in that way (for me personally).
Understanding read vs. write permissions, and only granting the access you actually need, helps too. Watching Claude Chrome navigate websites and take actions is such a wild learning experience too, highly recommend checking it out. Feel like it’s okay to babysit it until it gets more reliable/advanced.
27. What are your recommendations for getting buy-in for projects like the ones you shared, when AI is centralised and controlled by IT?
Demonstrate ugly v1s of AI solving super specific, valuable tasks (that stakeholders care about) using your own AI tools. Sales example: Want to better identify specific job title leads on LinkedIn? Here's a Claude Cowork automation I built that connects to Claude in Chrome to find and identify their LinkedIn profile urls in a spreadsheet.
My other tip is simply to make other people (usually stakeholders) look good. What kind of insights, customers, review data, competitive data could you surface to make those individuals look really sharp. +That does nothing but help get buyin + develop you as an AI pro on the team.
28. How relevant is traditional backlink building for AI visibility? Do backlinks play any role in building brand authority signals for AI chatbots?
Maybe a little, if they're supporting SEO; which is the primary way URLs get cited in AI (AI doesn't store URLs; any URL you see in an AI output is there because a search engine fed it that info in the background). Brand mentions are the new backlinks. I’d personally prioritize mentions over links today.
Category
Contact us if you have any suggestions on resources you would like to see more of, or if you have something you think would benefit our members.
Get in TouchSign up to receive updates on events, training and more from the MA.
.svg "MA-Logo (1)")