One of the reasons AI can sometimes feel daunting is the way it is discussed as a single entity, rather than a collection of specialised models. In this guest article, Calli Goldstein, Global VP, Customer Success at Pencil, argues there is no single “best” model that everyone should be using, and true transformation starts with encouraging experimentation with different models.
If there’s one thing I’d change about the way our industry talks about AI, it’s how it’s discussed as singular.
“AI” has become an umbrella term that attempts to cover the intricacies of every model and every use case. To me, this creates a conversation that feels intimidating. We’re essentially asking marketers to grapple with one enormous concept, rather than helping them understand the many practical ways these technologies can support the work they already do.
It’s also why discussions around AI tend to swing between two extremes. Either it’s spoken as a magic wand that will transform every part of marketing overnight, or as a negative force of disruption that will impact humanity for the worse.
Neither reflects the reality that marketers are experiencing – which is far more nuanced, and much more exciting.
AI isn’t one singular thing and it’s not going to change everything all at once. It’s a constantly growing collection of specialised models that individually create different solutions, for different problems, for different people. As one of our Creative Educators always says, “No one went to design school to specialise in resizing assets.” AI can operate as an assistant so the human in the loop can spend more energy on what matters, which is a new craft to learn in and of itself.
Once you start looking at AI through this lens, it stops feeling like one disruptive force and starts looking much more like a toolkit, where every marketer can find something that improves the way they work.
Start with your biggest pain point
One of the biggest mistakes businesses make is assuming they need to identify the perfect AI model before they’ve even worked out what they’re trying to achieve. Successful adoption starts with identifying the task that consistently slows someone down and asking how AI might make that process easier and faster.
That’s why I always encourage marketers to start with their biggest pain point rather than someone else’s. Every role across a marketing function has different pressures – a strategist might use an AI agent to brainstorm ideas or structure thinking before building their own approach, whereas a copywriter might use models like Claude or ChatGPT to explore multiple creative directions before refining the strongest one. Designers might find the greatest value in image and video generation models such as Veo 3.1 or Sora 2, which remove hours of repetitive production work, while someone working in research or planning could lean on tools like GWI’s AI-powered insights to synthesise large volumes of information more efficiently. Media teams, meanwhile, might use AI to analyse campaign performance and optimise creative in real time. The point isn’t that one model is better than another, but that each one excels at solving a different problem.
This requires a critical mindset shift, from “these tools will replace my expertise”, to “harnessing these tools enable me to become more efficient, effective, and evolve in my role.” And while there is a learning curve to implement new tools in existing systems, the payoff is in dividends as the human in the loop can release repeatable, time-consuming tasks to the LLM and focus on the expertise that add the greatest value, taste and judgement.
The key to this is a willingness to experiment. Marketing teams are under pressure to adopt AI quickly, and there is still a tendency to believe in a single “best” model that everyone should be using.
Teams should be willing to get comfortable with being uncomfortable. Try different models, compare what each one does well and see how they fit into day-to-day work. AI doesn’t need to be another enterprise software purchase that requires months of planning before anyone can use it.
I often think about it in terms of how children approach new things. They’re curious, quick to experiment and happy to play with the latest trend without overthinking it. Marketers would benefit from approaching AI in much the same way because nobody has all the answers yet, and some of the biggest breakthroughs still come from experimenting.
Making AI work across teams
We’re reaching a big shift in the way different AI assistants work together across an organisation. Agentic AI allows businesses to connect specialised assistants into broader workflows, with each one supporting a different stage of the marketing process before handing work seamlessly to the next.
Automating repetitive work creates significant time and cost savings, but the most interesting outcome is what people choose to do with the time they’ve gained. What will become of the increased capacity to challenge and improve the work, allowing them to focus on strategic thinking or creative judgement rather than execution alone? This can unlock a new paradigm of possibilities.
It also changes how knowledge moves around an organisation. Traditionally, the experience of senior strategists or creatives has been difficult to scale because so much of it exists inside people’s heads. Agentic workflows create an opportunity to capture that expertise, turning proven frameworks and ways of thinking into reusable systems that help junior team members learn faster and produce stronger work with greater consistency, de-siloing teams in the process.
AI represents change, and change can feel daunting. But we don’t expect children to master something new the first time they try it. We encourage them to stay curious, keep experimenting and build confidence as they learn. Businesses should give themselves that same permission.
So, it’s time we stopped talking about AI as if it’s one big external force that will transform business overnight. True transformation is incremental, unlocked by solving one workflow at a time, encouraging people to experiment and then connect those improvements across teams. It’s a more practical way to think about AI adoption, and ultimately a much more human one too.
