The Conditions That Turn AI Pilots Into Enterprise Value

This article is sponsored by HTEC and was written, edited, and published in alignment with our Emerj sponsored content guidelines. Learn more about our thought leadership and content creation services on our Emerj Media Services page.

AI adoption is rising, but the workflows that generate ROI remain largely unchanged — a pattern indicating that most deployments expand activity rather than impact.

According to the U.S. Census Bureau’s Business Trends and Outlook Survey, overall AI usage among American employer businesses hovered between 17% and 20% between December 2025 and May 2026, with 20% to 23% of businesses expecting to use it in the next six months — a pace of stated intent that consistently outruns actual deployment.

The gap is not one of ambition but of workflow design: Census Bureau researchers found that 57% of adopting firms integrate AI into three or fewer business functions, concentrated in sales, marketing, and strategy — evidence that most deployments never touch the operational core where ROI is generated.

Stanford University’s Institute for Human-Centered Artificial Intelligence (HAI) reports in its 2026 AI Index that while generative AI is now used in at least one business function at 70% of organizations, AI agent deployment remained in the single digits across nearly all business functions — the clearest evidence that experimentation has scaled while production-grade adoption has not. Despite commanding the largest share of global AI investment, Stanford HAI further notes that the United States ranks just 24th globally in AI adoption, at 28.3%, undercutting the assumption that capital and compute automatically translate into organizational readiness.

Together, these figures describe an enterprise landscape where technical capability has outpaced problem definition, workflow redesign, and change management. The result is a widening divide between AI that is piloted and AI that is productive — a gap rooted not in model performance, but in the absence of a defined business problem, a mapped human workflow, and a measurable adoption target before deployment begins.

In a recent Emerj series on the conditions that allow enterprise AI to deliver real business value, Emerj interviewed leaders across HTEC’s strategy, product, and technology organizations. Included in the series was a conversation with Carsten Wierwille, Chief Product and Design Officer, and Darko Todorovic, Chief Technology Officer, examining the upstream design, organizational, and measurement factors that determine whether AI initiatives advance beyond early promise and translate into durable enterprise impact.

This article examines four insights that clarify why enterprise AI initiatives stall and what conditions allow them to produce measurable, repeatable business value rather than isolated technical wins:

Guest: Carsten Wierwille, Chief Product & Design Officer at HTEC

Expertise: Digital Product Strategy, Product Design & Engineering, AI-Enabled Innovation, Business Transformation

Brief Recognition: Carsten Wierwille is a technology and product executive with more than 25 years of experience leading design, product, and engineering organizations globally. He currently serves as Chief Product & Design Officer at HTEC, where he leads product, design, and AI-focused consulting teams helping organizations develop digital products and technology solutions. He is also CEO of Momentum Design Lab, an HTEC company focused on digital product innovation and human-centered experiences. Previously, Carsten served as Global CEO of ustwo, a digital product company, where he led the transition from founder-led to employee-led ownership and helped establish the company as a B Corp. He has also held board and advisory roles with technology companies and has experience across business development, customer experience, product strategy, and organizational growth. Carsten holds a Master’s degree in Political Science from the University of Hamburg and studied at Indiana University Bloomington as part of a Ph.D. program exchange.

Problem Definition as the First Gate of AI Value

Carsten Wierwille describes problem definition as the single most important determinant of whether an AI initiative ever produces measurable business value. In his view, enterprises fail not because the models underperform, but because teams begin building before they understand the work itself — the workflow, the constraints, the user behavior, and the business outcome they are trying to change.

The absence of upfront problem definition is the root cause of wasted cycles and misaligned outputs, as Wierwille sees it:

Wierwille’s insight offers a practical tool for enterprise leaders: treat problem definition as a gating mechanism rather than a brainstorming exercise. In practice, this means requiring three artifacts before any AI work begins:

These artifacts force alignment across product, engineering, and operations, and they prevent the common enterprise failure mode Wierwille highlights: building before understanding.

The implication is simple but non‑negotiable: AI value is determined before a single line of code is written.

Organizational Readiness as the Driver of AI Scalability

Carsten Wierwille frames organizational readiness as the dividing line between AI that works in a pilot and AI that works in the enterprise. Pilots succeed because they are run by experts — people who already understand the workflow, the exceptions, and the judgment calls required to keep the system on track. Their competence absorbs friction. Their intuition fills gaps. Their experience compensates for what the AI cannot yet do.

But Wierwille argues that this dynamic creates a dangerous illusion of maturity. When leaders assume expert‑driven success will translate to the broader workforce, they underestimate the behavioral and operational shifts required for scale. Enterprise value emerges only when non‑experts — the majority of the organization — can adopt new workflows and decision patterns without slowing down, second‑guessing themselves, or requiring constant support.

According to Wierwille, the real test of readiness is whether the workflow works for everyone, not just the experts:

“Pilots give you a false sense of confidence because the people running them are already great at the job. They know how to correct the system, interpret edge cases, and keep the workflow moving even when the AI is imperfect. But that’s not scale. Scale is when someone who isn’t an expert can use the AI without friction, hesitation, or needing a specialist next to them. If the workflow only works for experts, it’s not ready for the enterprise.”

— Carsten Wierwille, Chief Product and Design Officer, HTEC

Wierwille’s insight gives leaders a practical lens: organizational readiness is the multiplier on AI value. It determines whether a workflow redesign becomes a company‑wide capability or remains a localized experiment. Readiness requires clear role expectations, training that matches the new decision model, and operational guardrails that make adoption safe for non‑experts — the conditions that turn pilot‑level success into enterprise‑level impact.

Guest: Darko Todorovic, CTO at HTEC Group

Expertise: Enterprise AI Strategy, Technology Leadership, AI Infrastructure & Engineering, Software Delivery

Brief Recognition: Darko Todorovic is a technology executive with more than a decade of experience leading engineering organizations, research initiatives, and enterprise technology delivery. He currently serves as Chief Technology Officer at HTEC, where he oversees engineering and delivery operations supporting enterprise AI initiatives across industries. Prior to becoming CTO, Darko held multiple leadership roles at HTEC, including VP of Engineering and Delivery, Director of Engineering and Delivery, and Head of R&D, helping scale the company’s technical capabilities and engineering organization. He has a background in electronics, control systems, and applied research, with doctoral studies at the Faculty of Electronic Engineering, University of Niš focused on haptic devices for minimally invasive surgery. Darko has also served as a teaching assistant at the University of Niš and contributed to technical education and research initiatives.

Cognitive Design as the Trust Layer for AI Output

Darko Todorovic frames cognitive design as the missing discipline in most enterprise AI deployments — the layer that determines whether users will trust, validate, and act on AI decisions once interfaces become automated. In his view, enterprises focus heavily on model performance but rarely define the criteria for what makes AI output trustworthy in production. Without those criteria, users either over‑trust or under‑trust the system, and both behaviors create operational risk.

Todorovic argues that cognitive design is not a UX exercise but a decision‑engineering function: it specifies how an AI system communicates uncertainty, how users are expected to interpret that uncertainty, and what validation steps must occur before action. When those expectations are not explicit, trust becomes inconsistent across teams — and inconsistent trust breaks workflows.

Enterprises must treat trust as a designed artifact rather than an emergent property:

Todorovic’s perspective offers a practical tool for enterprise leaders: codify trust criteria before deployment, not after adoption stalls. In practice, this means establishing three components:

These components ensure that trust is consistent, operational, and auditable — the conditions required for AI outputs to be relied on in production.

ROI Clarity as the Anchor of AI Initiatives

Darko Todorovic frames ROI clarity as the stabilizing force behind any enterprise AI initiative — the element that prevents drift, scope inflation, and the slow erosion of stakeholder confidence. In his view, AI efforts lose momentum not because the technology underperforms, but because leaders cannot articulate the business outcome the system is meant to change. Without that anchor, teams measure activity instead of impact, and progress becomes impossible to evaluate.

Todorovic emphasizes that ROI clarity is not a financial calculation performed at the end of a project; it is a design constraint established at the beginning. It defines the workflow that must change, the decision that must improve, and the metric that must move. When those elements are vague, teams build features that look impressive but do not shift the underlying business outcome. When they are explicit, the workflow redesign becomes obvious — and the AI system has a clear path to proving value.

Todorovic states that ROI clarity is what keeps AI work grounded in business reality rather than technical ambition:

Todorovic’s perspective gives leaders a practical discipline: anchor every AI initiative to a measurable business outcome before development begins. In practice, this means establishing three commitments that keep the work aligned:

These commitments ensure that AI efforts do not drift into feature‑building or experimentation. They keep teams focused on the business outcome, the workflow that drives it, and the measurable proof that value has been created.