HSD's Origin Story

Envisioning a new leadership role in data, AI and ML

The Problem: AI/ML projects are estimated to fail 85% of the time.

Our Perspective: We think it's way too high. At the same time, we know you can't opt out. The power of these tools - when you get it right - makes them essential investments. You need a partner who protects that investment from the unnecessary risk so you can keep taking the big swings that your market requires.

Our Solution: HSD is here to reduce that failure rate. We assess and fortify the common pitfalls that drive that 85%. Think of us as quality control, enabler and mitigator. HSD provides the risk assessment and mitigation to protect your investment in AI/ML, because we believe in its power to transform your brand and we want to see you win.

_____________________________

Patterned off an industry that gets it right

Early in my career, I worked in precision manufacturing. The company I worked for embraced high standards as a culture and an offering. Precision was their real product.

To hit this mark, they had a large quality department and those people were treated like judge, jury and executioner. They were autonomous - reporting directly to ownership.

They had full rights to say if something was fit to ship - or not. Profit be damned.

Their performance was measured differently than everyone else. While the business was measured on things like hitting ship dates and getting enough volume out the door, Quality was measured by the absence of failures after the product shipped. This meant that quality checks occurred throughout the entire process of making a product.

Even the purchase order for raw material had to have a quality stamp on it before it could be issued - lest some office worker decide to get the cheaper steel that unknowingly buckles under the heat this particular steel needs to withstand. A mistake like that may not be caught until the final stages of machining, or not at all. Catastrophic failure isn't an exaggeration.

When all your cost is invested in the product up front, it isn't enough to catch every error - you also want to catch it as early as possible to avoid compounding the sunk cost. And when your industry has compliance regulation (as do all customer and employee data projects) - delivering a faulty system carries heavy fines and worse.

This checks and balances system may seem like it would frustrate and slow down operations. But the opposite was true. It was less of an antagonistic relationship and more of a cooperative, symbiotic one. Because this wasn't some big brother thing, it was understood that the customer wasn't just buying a product that showed up on time, it had to work.

Even the departments measured by staying on-budget and on-time understood that shipping a bad product was lethal to the whole system.

It was worth having someone double check the work because there were no winners when bad product ended up in the hands of the customer.

______________________________

Adapting the template to the data industry

Unlike manufacturing, we aren't given exact specs and blueprints. These projects are complex and include a lot of 'art' in the process, introducing more subtle forms of failure:

• exploding cloud and infrastructure costs

• reputational risk of poor model outcomes (hallucinations and silly recommendations)

• emerging and changing compliance risk

• lackluster performance when recommendations aren't strong or the business fails to adopt it fully

Considering these risks, we believe the actual failure rate is much higher than 85%.

HSD is a firm dedicated to ensuring these projects don't just launch, but that they deliver the promised ROI. The manufacturing template is a great starting place but we know our work has to be more nimble and focused on making the right concessions and tradeoffs.

Too often, decisions are made without even realizing the downstream impact. We assess the common areas were this occurs and assist your teams in making more informed decisions. On the other hand, we just as commonly see development teams go out of their way to build complexity where it wasn't needed because the data scientist can adjust the model to meet the legitimate limitations of the system. Over-designing the system is just as a big of a risk as over-simplifying it.

Today our leading success metric seems to be "launch on time." This results in a number of unchecked, unbalanced risks. And it isn't anyone's fault. Because if time is all that matters, then launching on time is an unequivocable success. No matter what happens next.

And usually, its what happens next that drains resources, destroys trust and dries up the next investment in data and technology.

HSD takes on the missing link - serving as the expert who is focused on ensuring the project is profitable. When these projects go well, they go really well and additional tech investment is the natural result of that. We believe that ensuring quality and readiness of the project early is the tide that rises all boats.

In the end, there are very few situations that can't be overcome. It just takes having the right people at the table and we believe that HSD's role at that table is the missing link.

Our Why

We do this work because we know the power of these tools and we want to see them succeed. HSD was started by a data scientist after 15 years of enterprise leadership spent building and implementing AI/ML models. After witnessing the incredible returns that are possible when you get it right - there's nothing more heartbreaking than seeing the incredible losses when you get it wrong. We believe in this work and we want to see it succeed. Realizing those success stories is what drives us.

Learn more about How We Work and What We Focus On.

Or Contact Us to see how we can help.