AI from content to orchestration: How enterprise leaders are driving real transformation

As AI technology matures, a stark divide has emerged between companies leading with production-grade AI and those still trapped in pilot purgatory. Writer CEO May Habib and Templafy CEO Jesper Eriksen joined LinkedIn’s Tanya Dua to unpack what’s working — and what’s not — as enterprises attempt to operationalize generative AI.

Key takeaways

• Most enterprises are struggling to move from AI experimentation to impact due to issues with architecture, change management, and alignment on business logic.
• Leaders who focus on mission-critical use cases — those tied to revenue — are seeing the fastest ROI.
• The winning AI strategies are end-to-end, domain-specific, and designed with adaptability and orchestration at their core.

These insights came from our ScaleUp:AI event in November 2024, an industry-leading global conference that features topics across technologies and industries. Watch the full session below:

Most enterprises are still stuck at the starting line

Despite the hype, most enterprises are struggling to scale genAI solutions. “There is a real gulf between the folks who are still struggling to get to square one and the enterprises who are at square 10,” Habib said. She described many deployments as “in crisis,” not because the LLMs themselves are lacking, but because they’re not well integrated into real workflows. “Architecture matters,” she said. “Even state-of-the-art LLMs need to be connected to customer logic and use cases and data and workflows, and all of that needs to be maintained.”

The human side of AI: Change management is critical

For Eriksen, the greatest barrier isn’t technical — it’s psychological. “They are, with our help, rolling out a tool to a huge number of people who probably read in the newspaper that this is going to end their job,” he said.

The key is to approach AI adoption with empathy and communication. Templafy, which helps enterprises generate high-value documents, introduced an AI assistant to all 4 million of its users to encourage experimentation. “Start simple… small steps, but fast steps towards the kind of automated future of work,” he advised.

Build for the business, not the model

One of the biggest pitfalls enterprises face is focusing on model selection rather than business needs. “The biggest mistake folks make is analysis paralysis,” Habib warned. “You end up building a model museum.” She outlined four key focus areas that have nothing to do with the models themselves: mission-critical use cases, aligned business logic, real-time data pipelines, and organizational readiness to drive change.

Writer’s approach has been to create a family of domain-specific models — tailored for industries like finance, healthcare, and customer support. This focus reduces the need for fine-tuning and accelerates time to value. “[It’s] much more effective to be precision-trained for the enterprise, with a much more economical total cost of ownership,” she said.

Orchestration is the new integration

Eriksen emphasized the power of orchestration in delivering results. In Templafy’s case, that means automating every layer of document creation — from legal language and brand compliance to the insertion of real-time company data. “We can generate the perfect prompt… and send it to the right LLM who has the expertise to give the best answer,” he said

Revenue first: The real measure of readiness

For both leaders, identifying which use cases to pursue begins with one question: Will it move the top line? “You can’t take productivity to the bank,” Habib said. Writer has helped consumer goods companies cut time-to-market in half and retailers increase online revenue by more than 20%. “Start where we already have alignment on how this thing should be done… that gets you to value so much faster,” she advised.

Cutting through hype, avoiding tech debt

With the flood of AI solutions on the market, it’s easy for enterprises to fall into “tech debt” or vendor lock-in. But the panelists agreed that today’s AI software isn’t as rigid as legacy systems. “Switching costs are actually way less than they ever were,” said Habib. The key is to invest in capabilities that build internal resilience: use-case clarity, quality data, skilled people, and the appetite to change.

What it takes to win

In a saturated market, standing out requires more than clever demos. “We’re not just a point solution,” Habib said. “Enterprises are betting on innovation roadmaps — and we’ve proven we can deliver.” Eriksen echoed the focus on long-term value: “There’s a technology roadmap, but in the end, it’s a value creation—for you—roadmap. You will generate documents without manual interaction that generate more revenue to you as a company.”