Asking the right questions about AI: How we helped a client achieve success

/in /by

While other, larger institutions haven’t seen the promised benefits of AI, our client Historic Images has achieved what their leaders knew was possible with this technology.

That’s because they were willing to find the right partner and ask the right questions before deciding on a solution.

Historic Images digitizes and preserves massive print photo archives from newspapers, media companies, and other organizations. This allows the owners of those archives to unlock new value from images that would otherwise sit unseen in filing cabinets, or that could have been lost or destroyed.

For years this work relied on manual effort. After scanning the images, teams of people reviewed and interpreted the content, then added the metadata required to make those images searchable and to give them context and value – dates, places, descriptions, and people’s names.

This manual approach worked, but only up to a point. When they came to us, Historic Images was processing about 500,000 images per year. To reach their growth goals, they needed to increase that output to 10 million images per year.

Leaders at Historic Images understood something important that many teams don’t realize. Yes, you can refine workflows, optimize steps, and improve coordination, but when the gap between your current performance and your vision of success is too big, incremental improvements alone aren’t going to close it.

They also knew that AI could probably address their concerns, but they didn’t know where to start.

How do you know AI is the right solution?

As Jeremy Upton, our director of project success, explained in a recent presentation, even though Historic Images came to us with AI in mind, we didn’t assume that a custom AI solution was the best answer.

Instead, Jeremy said, Skeleton Key focused first on understanding the nature of the work itself and asking key questions:

  • What problem are we solving, and what’s the root cause of that problem?
  • What do we hope to achieve by solving it, and how will we measure success?
  • What possible solutions will deliver the most value, quickly?

Once we understood what was required, the constraints they were facing, and what kind of results would actually create value for the business, we were ready to make recommendations.

In this case, the challenge wasn’t just a matter of volume, but the need for interpretation. Basic image recognition tools can identify general elements within a photo, but that level of detail doesn’t provide much value in a business context. Recognizing that an image contains a person or a building is very different from identifying who that person is, what event is being captured, or why the image might be historically or commercially significant.

This level of interpretation is where traditional tools fall short, and where large language models (LLMs) have a real advantage. And since Historic Images required both the ability to operate at scale and contextual understanding, we knew that AI wasn’t just an option worth considering. It was the best option.

The results speak for themselves

In that recent presentation, Jeremy and Greg Lane, our director of technology, give a technical overview of what we built for Historic Images. It’s worth your time if you’re considering an AI solution and want to see how the different elements function, and how we approached building a custom LLM workflow.

Jeremy also shares the thought process of how we answer those question in discovery with a current or potential client. And that’s a very different approach than you’ll find from many solutions vendors. A lot of B2B software and AI providers will promise you the Moon and then deliver green cheese, all while assuring you that they’re listening to your feedback.

But because we asked the right questions up front, today Historic Images has reduced their image processing costs by 98 percent, with improved data quality and reliability as well. We also built them a custom software interface and an improved storage solution to help them work more efficiently.

This success was possible because we worked with their team to understand their needs, challenges, and constraints, then create a solution tailored to them. AI was an important part of that solution, but we didn’t take that for granted when we started the conversation.

Is your vendor helping you ask the right questions?

If you assume that AI is always going to be the answer to your problems, you risk applying it in situations where it introduces unnecessary complexity or cost. But if you start with a clear understanding of your processes and the problems you’re trying to solve, you’re much more likely to find the right solution. In many cases, a custom software solution or AI workflow can deliver more value than an off-the-shelf solution that forces you to change how you work.

If you’re considering an off-the-shelf AI tool or SaaS solution, but aren’t convinced it’ll fit your needs or deliver the value it promises, it’s worth your time to watch the presentation with Jeremy and Greg and see how we approach this type of challenge.

And if you’d like to find out how we could help your organization with a specific software or workflow challenge, schedule a call with us today – no sales pitches or demos, just a straightforward conversation.