Horwood describes herself as a marketer with a strong technology and data background, having spent twenty years in digital roles including chief digital officer, which put her at the forefront of AI applications well before generative tools arrived.
Her approach is deliberately near term: take the native tools the company already has, use the AI features inside them, and get as many people using them as possible, because the use cases will come from those people rather than from her.
Key findings
- About 95 percent of the employee population has access to the tools and uses them regularly, after a year and a half of extensive training including large-scale events the company calls AI discovery days.
- The chief executive’s endorsement mattered but did not do the work. Horwood notes that just because the CEO says something does not mean everybody jumps on board and uses it.
- The first marketing use case is scheduling. The company puts out 20,000 regulated pieces a year, and materials deadlines and key dates that were still held in spreadsheets have been put into a model that prompts marketers about when to start.
- The second is briefing. The team codified what a great brief looks like and is building a system that assembles a brief from the right insights, brand systems and research at the press of a button, against a volume of around 300 briefs.
- Her position on jobs is that AI is not going to replace your job, but not knowing AI will.
Choosing projects that finish
Horwood is explicit that she picked work the commercial organization can complete on its own power. Bigger projects move too slowly, and she is not working on moon shots.
Her operating guideline is that people need to use the tools and put them to work, using AI daily and for their actual work, because that is how they come to understand what is possible. She focused almost entirely on things that help in the next year or two.
The C-suite partners are fewer than expected. She works with IT and, more than people would think, with finance, because automation costs money and finance is looking for savings everywhere.
Why legal is not the obstacle it sounds like
In a highly regulated industry, the constraint turns out to be manageable because the tools were legally vetted before deployment. Horwood describes it as freedom within a framework.
The second reason is that she is automating processes that are already approved, so the legal review is not a heavy lift. She notes there are very few examples where it is.
She is candid that compliance, regulatory and legal groups did question whether putting the tools in every employee’s hands was a good idea, and that this took discussion.
Narrowing fifty ideas to three
An innovation team inside her group ran an idea day with Microsoft a year and a half earlier, at a point when people were not sure what AI could do for them. Teams pitched what they wanted to accomplish and the partner evaluated the ideas in real time.
Horwood calls the process very educational, partly because teams were told when AI was not the answer, or when another tool would solve the problem. The list went from around 50 ideas to 10, and then a top three, much of it about democratization of information and insights.
No AI office, and what the human is for
Horwood draws a direct parallel to digital. Companies wanted a digital office so somebody else would handle it, and she believes the same thing is happening with AI. Her company has not created an AI office, and instead empowers individual employees and teams through role based training, because the tool is used very differently by an experience optimization role and a marketing strategy role.
On the tension between human and machine, she describes designing programs to make AI an accelerant rather than a replacement, and insists judgment remains. Her line on creative work is that just because AI can make an ad does not mean it is good.
The change she describes is uncomfortable rather than threatening. What is taken out is the work humans do not need to do, and she acknowledges people dislike that, because someone whose day is spent assembling things derives real value from being the only person who can. Her reframe is that the job is now thinking about what you are assembling.
Her definition of the human job is synthesis and storytelling, and the point of the automation is to create the conditions to have more time for simplification and storytelling. She adds, dryly, that it sounds like something nobody would refuse, but that some people do.
What she still wants measured
Asked what she would want from the study, Horwood raises the measurement gap. She does not believe anybody is really measuring whether increased time on thinking, storytelling and synthesis actually contributes to the bottom line, and says those are the things that need to be looked at.
The other area she points to is segmentation, where the company uses AI to uncover untapped need, with the specific example of segmenting people who are undiagnosed for a disease, bringing in more variables than was previously possible.
On what marketing is doing with AI day to day, she describes it as embedded rather than separate: taking it into the workflow to gain better behavioral insights, design stronger experience plans, and optimize those plans and create triggers. She does not make a show of calling it an AI initiative, because she just calls it modern.
Scope and limitations
This is a single interview with one commercial leader responsible for the US market, and she says so directly. The account covers the commercial side of the business, not the medical research and drug development side, which she describes as already heavily invested in large language models and outside her remit.
Source
This page summarises The Transcript: Interview with Gail Horwood (Novartis), by Mariana Peneva, Gail Horwood, December 5, 2024.