Microsoft CCO Frank X. Shaw did a fireside chat [full video] at the Wall Street Journal Leadship Institute CMO Council Summit where he explained in depth how corporate communications can use and benefit from AI. The main thrust of his argument is that generative AI shouldn’t replace human storytelling. Instead, it should strip out the “mechanical junk” so humans can spend more time on the deeply human parts: finding great stories, falling in love with them, shaping them, then sanity-checking how they land with real people.
“AI as power tools, humans as architects.”
He made lots of great points, many of which are integral parts of our CommsTransform™ IDEA framework such as AI as a transformer, AI as a personal coach, AI as cultural change and ensuring AI is grounded in the right knowledge and data.
but the two that stood out to me were two ideas I already use extensively and help clients to use.
AI as a transformer
Frank explained how it is up to the human to find the story, have the great idea and create the excellent “core document”. Then AI can be used to transform it.
AI does the drudgery of turning it in to an internal podcast for sales, a script for a TikTok or Insta Reels explainer, a Q&A for field teams or talking points for executives.
AI can help with different versions for different audiences. A peer-reviewed paper for academics or a technical white paper for scientists. It can help with adapting the core document for multiple channels – landscape and portrait video, an audio podcast, a PowerPoint presentation, a ‘listicle’ and more.
The old adage of “Garbage In, Garbage Out” applies so it’s critical that the original core document is actually excellent.
AI as a critic and coach
One of my favourite and most productive uses of AI is as a constructive (or even destructive!) critic and personal coach or advisor. You keep the human in the loop by actually writing the news release, speech or white paper yourself, but then using AI to interrogate it:
- Act as our toughest competitor and attack this.
- Act as an experienced, informed and sceptical BBC journalist and ask the toughest questions possible.
- Ask AI to identify the three top points to see if they match the top messages you intended to convey.
The last is particularly important in the new world where PR is trying to influence AI answers using GEO/AEO techniques. If you ask your own AI to summarise it and it doesn’t identify the most important messages, then your story is unlikely to work its way effectively into AI answers or overviews.
AI for crisis communications
Frank used the example of the Crowdstrike outage to illustrate how powerful AI can be (short clip on LinkedIn). It wasn’t Microsoft’s fault but Microsoft’s blue screen of death is what users saw in front of them so naturally blamed Microsoft. Microsoft’s reputation was at the front of the frame, even if technically it wasn’t anything to do with Microsoft.
The Microsoft corporate communications team used Meltwater and an AI agent in Teams to track global coverage in near real time and analyse how the story was framed to see if it was “Microsoft is broken…” or “Crowdstrike broke…”
Using the AI agent, it assessed the “blast radius” to see which channels the story was spreading fastest. This enabled them to decide it was a “broadcast story” on TV, radio and big news outlets rather than just online and social media.
It helped dictate the response of getting real “Microsoft humans with blue badges” on to TV and radio as quickly as possible to inject real facts without just blaming Crowdstrike.
AI was used to inform the humans. It was still human judgement, expertise and experience that made the decisions about what to do. They just did so from a position of greater knowledge.
AI is cultural change, not just technology
A lot of our work at Purposeful Relations is about unlocking the AI Adoption Illusion. This is where companies provide AI tools and expect miraculous transformation, only to see disappointing results.
Frank had several ideas and thoughts around this. His first point was to not think of AI as a mysterious monolith to be worshipped
Another point he made strongly mirrored something we encourage which is for different people to use AI in different ways to suit their personal likes, dislikes, strengths and weaknesses.
Good writers who enjoy writing are unlikely to use AI for drafting. I’m writing this from scratch using an AI generated transcript of the video. When my draft is finished I’ll get AI to critique it and check it. However the keen writer might use AI to help turn their article into a PowerPoint presentation as they are neither good at it or enjoy it.
AI as a good habit
He also spoke about the importance of “habit-building” and used the example of a daily prompt email. Every morning, everyone in the team gets a single prompt to try. Sometimes they are practical like “Summarise my inbox by topic” and sometimes fun such as “Analyse my inbox as if is Star Wars – who is Yoda, who is Darth Vader.”
I can confirm that the Star Wars one is fun and astonishingly astute.
The daily prompt is designed to help overcome fear and reluctance as well as provide ideas for creative uses for AI. It gets everyone using AI daily to make it habit-building in low-risk, fun ways.
AI as a team sport
Many people’s first experiments with AI are for personal use and then they’ll start using the same tool for work (usually without permission or telling their boss – Shadow AI – which is extremely risky). However, AI’s real benefits are when used across a team.
This moves AI from individual productivity to team productivity. Frank uses the example of publishing a story to the corporate website. This can be broken down into multiple discrete steps. For each step ask “How can AI help here?”. Test AI on each step to see how it helps.
This enables you to improve and redesign team workflows.
AI gets better so keep iterating
It’s not just next generation models that improve AI, but also new features and tools within existing models. This means it’s critical to keep monitoring and understanding AI improvements as steps, tasks or workflows that failed six months ago might now be viable.
My personal favourite example of this is design in Microsoft 365 Copilot. When I first attempted to use it, it was terrible (and that’s probably being too complimentary!).
AI accuracy and trust
When ChatGPT first launched it hallucinated a lot, which is a fancy way of saying it lied and made stuff up. Today’s AI models rarely hallucinate, but still aren’t perfect. People make the mistake of thinking of them as machines that can provide right or wrong answers. Instead, it’s better to think of them as human assistants where you still need to check their work. AI models have graduated from being work-experience interns to experienced junior managers.
AI based on your knowledge
AI tools are at their most useful when grounded in your companies own data, knowledge and expertise. AI tools using your own data sources are unlikely to hallucinate. That’s the single greatest strength of Microsoft 365 Copilot. It knows everything about your company (that you choose to give it access to).
A prompt in Copilot can safely contain private and confidential information or sensitive data as it is protected by the same enterprise-grade data security as your emails and files in SharePoint.
In fact AI knows more than any employee about your company because it can access so much more data than a human. This means you can create expert AI agents in Copilot Studio that are trained to base answers on internal reports, policies, meeting transcripts, briefings and knowledge bases,
This massively reduces hallucinations and mean answers reflect your companies culture, perspective and approaches rather than simply plucking ideas off the open web.
It also means AI outputs are auditable and you can identify the source of the outputs.
The reluctance of some IT teams to benefit from Copilot is often they’ve created what Frank X Shaw refers to as a “data swamp” where information is scattered across email, Teams, SharePoint sites and folders. Often users permissions are fragmented so people can’t access knowledge they need to, while others can see data they shouldn’t.
Frank advises draining the swamp to create a clean central repository AI can safely use as a reputable data source.
AI adoption at scale
These are some of the actions and approaches that are needed to ensure PR and communications teams unlock the full benefits of AI adoption.
Success doesn’t just mean improved productivity – time-savings of two-to-four hours a week are not unreasonable. Even greater success comes from unlocking greater human potential. As AI absorbs some tasks it empowers humans find the stories that matter, use critical thinking and judgement, spend quality time with real people and to concentrate on creativity and nuance,
Are you ready?
Doing all of this requires time and expertise. If you need help successfully developing and adopting AI then get in touch with me at Purposeful Relations to find out how are CommsTransform™ IDEA framework can unlock the potential of AI for you.
