Practical techniques for using AI to guide your thinking, validate sources, and build structured outputs.

The DVT Insights webinar on 17 April brought a practical look at how generative AI reshapes professional research. Anton Keyter, who heads up AI strategy and product development at DVT, led the session titled "AI as Your Research Partner: How to ask, explore, and explain better", sharing techniques to turn tools like ChatGPT, Grok, and Gemini into reliable collaborators for tackling complex questions and producing usable outputs.

Unlike marketing hype or theoretical discussions, this session focused squarely on how professionals, from strategists and analysts to business consultants and technical leads, can use AI to explore questions more thoroughly, verify findings, and clearly communicate their insights.

From reactive answers to research partnerships

Keyter began by mapping the progression of large language models from basic question-answering tools to more sophisticated systems capable of engaging in back-and-forth reasoning. "We're not just typing a question and hoping for a decent paragraph anymore," he says. "We're prompting the system to think with us, posing questions, refining directions, and delivering structured responses we can actually use."

That structure matters. AI outputs can now be shaped into summaries, draft reports, presentation-ready text, and even HTML-formatted documents freeing up professionals to spend less time writing and more time analysing. However, the usefulness of these models depends heavily on how they're used.

A live demo in deep research

The session's highlight was a live demonstration of what Keyter described as "deep research", a method that combines human critical thinking with AI's speed and pattern recognition. The topic? A pressing global issue: How are AI and geopolitical instability reshaping the future of work, and how should individuals adapt?

Starting with this open-ended question, Keyter walked attendees through how ChatGPT, when properly guided, could identify underlying themes, recommend source material, formulate further lines of inquiry, and present the findings in structured, referenceable formats.

The session showcased three platforms, ChatGPT, Gemini, and Grok, working in parallel. This multi-model approach allowed users to see not just one AI's take but how each tool interpreted the same prompt differently. Keyter noted that this process is key in reducing bias and confirming validity: "When multiple models converge on similar themes or sources, you gain more confidence in the consistency of the findings."

The art of the prompt

Central to the success of AI-assisted research is knowing how to prompt effectively. As Keyter emphasised, good outputs begin with good inputs. This includes:

• Asking the AI what questions should be asked before diving into a topic.
• Avoiding "leading prompts" that push the model toward a pre-decided answer.
• Requesting citations and then verifying whether those citations are accurate and traceable.
• Treating AI like a co-pilot capable of suggesting and refining ideas, but not driving solo.

This mindset also helps address one of the biggest challenges with generative AI - hallucinations. When asked for factual answers, some models fabricate sources or details. That's why validation remains essential. "Never copy and paste from an AI into a client-facing document without checking the references," Keyter advises.

Broad relevance across industries

While many associate AI tools with data science or software engineering, Keyter made it clear that their applications go far beyond technical fields. Marketing teams can use AI to create campaign strategies or product messaging guides. HR professionals can build internal communications or policy summaries. Sales teams can prepare tailored pitch decks. Academic and policy researchers can save time drafting working papers.

"This isn't a tool limited to coders or AI specialists," says Keyter. "If your job involves thinking, writing, or problem-solving, AI has something to offer."

In particular, ChatGPT Pro users now benefit from multimodal capabilities, allowing the tool to handle documents, spreadsheets, diagrams, and more. However, even those starting with free tools like Google Gemini can gain a significant productivity boost, provided they approach the process with a clear strategy.

Advice for getting started

For professionals exploring AI-assisted research for the first time, Keyter shared a few simple but effective tips:

• Start with free tools and get comfortable prompting.
• Keep sessions focused, and don't try to solve too much in one go.
• Use AI to brainstorm angles, generate outlines, and stress-test assumptions.
• Document your process. It helps you audit your work later and refine how you use the tools.

He also suggested setting up a basic prompt library, snippets that you can reuse or adapt to save time in future sessions. This might include prompts for summarising, questioning, formatting, or drafting.

The human role remains central

Despite AI's speed and convenience, the session repeatedly reinforced one core idea: AI augments thinking but doesn't replace it.

"These tools can help you move faster," says Keyter, "but you must still apply judgment. What you choose to ask, how you interpret the response, and how you structure the output - that's where the human value lives."

As more individuals experiment with generative AI in their daily workflows, organisations that develop internal training to help people use these tools thoughtfully will achieve the greatest benefits.

Watch the full webinar.