The Conditions That Turn AI Pilots Into Enterprise Value

 This article is sponsored by HTEC and was written, edited, and revealed in alignment with our Emerj sponsored content guidelines. Learn extra about our thought management and content material creation companies on our Emerj Media Services page.

AI adoption is rising, however the workflows that generate ROI stay largely unchanged — a sample indicating that the majority deployments increase exercise moderately than impression.

According to the U.S. Census Bureau’s Business Trends and Outlook Survey, total AI utilization amongst American employer companies hovered between 17% and 20% between December 2025 and May 2026, with 20% to 23% of companies anticipating to make use of it within the subsequent six months — a tempo of acknowledged intent that persistently outruns precise deployment.

The hole isn’t considered one of ambition however of workflow design: Census Bureau researchers found that 57% of adopting corporations combine AI into three or fewer enterprise capabilities, concentrated in gross sales, advertising, and technique — proof that the majority deployments by no means contact the operational core the place ROI is generated. 

Stanford University’s Institute for Human-Centered Artificial Intelligence (HAI) reports in its 2026 AI Index that whereas generative AI is now utilized in not less than one enterprise operate at 70% of organizations, AI agent deployment remained within the single digits throughout practically all enterprise capabilities — the clearest proof that experimentation has scaled whereas production-grade adoption has not. Despite commanding the biggest share of world AI funding, Stanford HAI additional notes that the United States ranks simply twenty fourth globally in AI adoption, at 28.3%, undercutting the idea that capital and compute routinely translate into organizational readiness. 

Together, these figures describe an enterprise panorama the place technical functionality has outpaced downside definition, workflow redesign, and alter administration. The result’s a widening divide between AI that’s piloted and AI that’s productive — a niche rooted not in mannequin efficiency, however within the absence of an outlined enterprise downside, a mapped human workflow, and a measurable adoption goal earlier than deployment begins.

In a latest Emerj sequence on the situations that enable enterprise AI to ship actual enterprise worth, Emerj interviewed leaders throughout HTEC’s technique, product, and know-how organizations. Included within the sequence was a dialog with Carsten Wierwille, Chief Product and Design Officer, and Darko Todorovic, Chief Technology Officer, analyzing the upstream design, organizational, and measurement elements that decide whether or not AI initiatives advance past early promise and translate into sturdy enterprise impression.

This article examines 4 insights that make clear why enterprise AI initiatives stall and what situations enable them to provide measurable, repeatable enterprise worth moderately than remoted technical wins:

• Problem definition as the primary gate of AI worth: AI solely produces measurable outcomes when groups outline the enterprise downside and human workflow upfront, as a result of constructing earlier than understanding results in wasted cycles and misaligned outputs.
• Organizational readiness as the motive force of AI scalability: Pilots succeed with specialists, however enterprise worth emerges solely when the broader workforce can undertake new workflows and choice patterns with out friction.
• Cognitive design because the belief layer for AI output: As interfaces turn out to be automated, enterprises should outline how customers interpret, validate, and act on AI selections, as a result of with out shared belief standards, outputs can’t be relied on in manufacturing.
• ROI readability because the anchor of AI initiatives: AI efforts drift when leaders can not outline the enterprise final result up entrance; with no baseline and a goal metric, worth can’t be measured or confirmed.

Episode 3:  Why the Way AI Feels Is as Important as How It Works – with Carsten Wierwille of HTEC

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 know-how and product government with greater than 25 years of expertise main design, product, and engineering organizations globally. He at the moment serves as Chief Product & Design Officer at HTEC, the place he leads product, design, and AI-focused consulting groups serving to organizations develop digital merchandise and know-how options. He can also be CEO of Momentum Design Lab, an HTEC firm centered on digital product innovation and human-centered experiences. Previously, Carsten served as Global CEO of ustwo, a digital product firm, the place he led the transition from founder-led to employee-led possession and helped set up the corporate as a B Corp. He has additionally held board and advisory roles with know-how firms and has expertise throughout enterprise growth, buyer expertise, product technique, and organizational development. Carsten holds a Master’s diploma in Political Science from the University of Hamburg and studied at Indiana University Bloomington as a part of a Ph.D. program alternate.

Problem Definition because the First Gate of AI Value

Carsten Wierwille describes downside definition as the only most vital determinant of whether or not an AI initiative ever produces measurable enterprise worth. In his view, enterprises fail not as a result of the fashions underperform, however as a result of groups start constructing earlier than they perceive the work itself — the workflow, the constraints, the person conduct, and the enterprise final result they’re making an attempt to vary.

The absence of upfront downside definition is the basis explanation for wasted cycles and misaligned outputs, as Wierwille sees it:

“Most AI tasks don’t fail as a result of the know-how is insufficient — they fail as a result of no person agreed on the issue within the first place. If you don’t perceive the workflow, the constraints, and the human selections inside that workflow, you’re not constructing an AI resolution; you’re constructing a guess. Teams rush into growth pondering pace is the benefit, however pace with out readability simply accelerates misalignment.”

— Carsten Wierwille, Chief Product and Design Officer, HTEC

Wierwille’s perception affords a sensible device for enterprise leaders: deal with downside definition as a gating mechanism moderately than a brainstorming train. In observe, this implies requiring three artifacts earlier than any AI work begins:

• An outlined enterprise final result — the measurable change the AI should produce.
• A mapped human workflow — the steps, selections, and constraints the AI will increase or automate.
• A person‑conduct mannequin — how individuals at the moment make selections and the way these selections should change.

These artifacts power alignment throughout product, engineering, and operations, they usually forestall the widespread enterprise failure mode Wierwille highlights: constructing earlier than understanding.

The implication is easy however non‑negotiable: AI worth is set earlier than a single line of code is written.

Organizational Readiness because the Driver of AI Scalability

Carsten Wierwille frames organizational readiness because the dividing line between AI that works in a pilot and AI that works within the enterprise. Pilots succeed as a result of they’re run by specialists — individuals who already perceive the workflow, the exceptions, and the judgment calls required to maintain the system on observe. Their competence absorbs friction. Their instinct fills gaps. Their expertise compensates for what the AI can not but do.

But Wierwille argues that this dynamic creates a harmful phantasm of maturity. When leaders assume professional‑pushed success will translate to the broader workforce, they underestimate the behavioral and operational shifts required for scale. Enterprise worth emerges solely when non‑specialists — nearly all of the group — can undertake new workflows and choice patterns with out slowing down, second‑guessing themselves, or requiring fixed help.

According to Wierwille, the true check of readiness is whether or not the workflow works for everybody, not simply the specialists:

“Pilots offer you a false sense of confidence as a result of the individuals working them are already nice on the job. They know methods to right the system, interpret edge circumstances, and preserve the workflow transferring even when the AI is imperfect. But that’s not scale. Scale is when somebody who isn’t an professional can use the AI with out friction, hesitation, or needing a specialist subsequent to them. If the workflow solely works for specialists, it’s not prepared for the enterprise.”

— Carsten Wierwille, Chief Product and Design Officer, HTEC

Wierwille’s perception provides leaders a sensible lens: organizational readiness is the multiplier on AI worth. It determines whether or not a workflow redesign turns into an organization‑extensive functionality or stays a localized experiment. Readiness requires clear position expectations, coaching that matches the brand new choice mannequin, and operational guardrails that make adoption protected for non‑specialists — the situations that flip pilot‑stage success into enterprise‑stage impression.

​

Episode 5:  How Enterprise Leaders Should Measure the ROI of AI – with Darko Todorovic of HTEC

​​

Guest: Darko Todorovic, CTO at HTEC Group

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

Brief Recognition: Darko Todorovic is a know-how government with greater than a decade of expertise main engineering organizations, analysis initiatives, and enterprise know-how supply. He at the moment serves as Chief Technology Officer at HTEC, the place he oversees engineering and supply operations supporting enterprise AI initiatives throughout industries. Prior to changing into CTO, Darko held a number of management roles at HTEC, together with VP of Engineering and Delivery, Director of Engineering and Delivery, and Head of R&D, serving to scale the corporate’s technical capabilities and engineering group. He has a background in electronics, management techniques, and utilized analysis, with doctoral research on the Faculty of Electronic Engineering, University of Niš centered on haptic gadgets for minimally invasive surgical procedure. Darko has additionally served as a instructing assistant on the University of Niš and contributed to technical schooling and analysis initiatives.

Cognitive Design because the Trust Layer for AI Output

Darko Todorovic frames cognitive design because the lacking self-discipline in most enterprise AI deployments — the layer that determines whether or not customers will belief, validate, and act on AI selections as soon as interfaces turn out to be automated. In his view, enterprises focus closely on mannequin efficiency however not often outline the standards for what makes AI output reliable in manufacturing. Without these standards, customers both over‑belief or below‑belief the system, and each behaviors create operational threat.

Todorovic argues that cognitive design isn’t a UX train however a choice‑engineering operate: it specifies how an AI system communicates uncertainty, how customers are anticipated to interpret that uncertainty, and what validation steps should happen earlier than motion. When these expectations aren’t express, belief turns into inconsistent throughout groups — and inconsistent belief breaks workflows.

Enterprises should deal with belief as a designed artifact moderately than an emergent property:

“If you don’t outline how individuals ought to interpret an AI choice, they’ll invent their very own logic — and that logic will range from individual to individual. Some will belief the system an excessive amount of, others under no circumstances, and each behaviors create failure modes. Trust isn’t one thing you hope emerges; it’s one thing you design. You need to specify what the system is aware of, what it doesn’t, and what the person is predicted to do with that info.”

— Darko Todorovic, Chief Technology Officer, HTEC

Todorovic’s perspective affords a sensible device for enterprise leaders: codify belief standards earlier than deployment, not after adoption stalls. In observe, this implies establishing three parts:

• Interpretation guidelines — how customers ought to learn confidence scores, uncertainty, or structured outputs.
• Validation steps — the required checks earlier than an AI‑assisted choice is finalized.
• Action protocols — the outlined behaviors customers should observe when the AI recommends, predicts, or escalates.

These parts be sure that belief is constant, operational, and auditable — the situations required for AI outputs to be relied on in manufacturing.

ROI Clarity because the Anchor of AI Initiatives

Darko Todorovic frames ROI readability because the stabilizing power behind any enterprise AI initiative — the aspect that forestalls drift, scope inflation, and the gradual erosion of stakeholder confidence. In his view, AI efforts lose momentum not as a result of the know-how underperforms, however as a result of leaders can not articulate the enterprise final result the system is supposed to vary. Without that anchor, groups measure exercise as an alternative of impression, and progress turns into not possible to judge.

Todorovic emphasizes that ROI readability isn’t a monetary calculation carried out on the finish of a undertaking; it’s a design constraint established at the start. It defines the workflow that should change, the choice that should enhance, and the metric that should transfer. When these parts are obscure, groups construct options that look spectacular however don’t shift the underlying enterprise final result. When they’re express, the workflow redesign turns into apparent — and the AI system has a transparent path to proving worth.

Todorovic states that ROI readability is what retains AI work grounded in enterprise actuality moderately than technical ambition:

“If you may’t outline the enterprise final result upfront, you’re not doing AI — you’re doing experimentation. Teams will construct attention-grabbing options, however none of these options will add as much as worth as a result of there’s no baseline and no goal. ROI isn’t one thing you calculate on the finish; it’s one thing you outline at the start. It’s the constraint that retains the work sincere.”

— Darko Todorovic, Chief Technology Officer, HTEC

Todorovic’s perspective provides leaders a sensible self-discipline: anchor each AI initiative to a measurable enterprise final result earlier than growth begins. In observe, this implies establishing three commitments that preserve the work aligned:

• A baseline — the present efficiency of the workflow or choice the AI will increase.
• A goal metric — the measurable change the AI should produce to justify deployment.
• A workflow speculation — the precise behavioral or operational shift required to realize that metric.

These commitments be sure that AI efforts don’t drift into characteristic‑constructing or experimentation. They preserve groups centered on the enterprise final result, the workflow that drives it, and the measurable proof that worth has been created.