AI adoption is now less a technology question than a management design problem. The article rightly points to architecture, trust and leadership fluency, but the harder issue for CIOs and transformation leaders is deciding who owns AI decisions, which risks require formal approval, and how much autonomy is acceptable by process, function and customer journey.
That means enterprises need a clearer operating model before they scale tools. A workable approach is to define decision rights for model selection, data use, human override, monitoring and exception handling. Without that discipline, AI pilots can multiply faster than accountability, creating inconsistent customer experiences, uncontrolled shadow adoption and unclear auditability. The practical question is not โcan we deploy agentic AI?โ but โwhere does the business tolerate machine-led action, and where must humans remain the accountable actor?โ
There is also a portfolio issue. AI initiatives should be ranked against measurable value pools, not enthusiasm or vendor pressure. Leaders should ask which use cases improve margin, speed, service quality or risk reduction enough to justify the data cleanup, integration work and change effort they demand. A narrow set of high-value workflows usually beats a broad spread of disconnected experiments.
Finally, adoption will depend on workforce confidence as much as technical capability. If AI is introduced as a replacement narrative, resistance rises; if it is framed as a redesign of work, with explicit reskilling and role clarity, uptake improves. Next questions for executives: which decisions will AI be allowed to make, who signs off on the exceptions, and how will benefits be measured after go-live rather than at pilot stage?
Whatโs holding companies back from taking full advantage of the technology, and which trends should they keep in mind?
AI is no longer just a nascent technology โ it is a business imperative. However, many organisations have yet to fully wrap their heads around AI to maximise its potential. According to Accentureโs Technology Vision 2025ย report, only 13 percent of enterprises see significant impact from AI at scale.
What are some of the critical roadblocks standing in the way of broader AI adoption, and what opportunities can companies take advantage of? These themes were the focus of a recent Tech Talk Xย webinarย I moderated, featuring Frรฉdรฉric Brunier, Accentureโs Global Lead for the Technology Strategy practice, andย Emmanuel Viale, a managing director in Accentureโs Technology Strategy & Advisory group.
Four key AI trends to watch
Brunier explained that this yearโs Accenture report explores the potential of AI โ in particular, generative AI (GenAI) and agentic AI โ to disrupt businesses and industries within a three-to-five-year period across multiple enterprise dimensions.
โWhat we clearly see is that AIโs adoption of human-like capabilities is significantly gaining in speed, and the most significant change with AI is its ability to learn, reason and make decisions autonomously,โ Brunier said. โBut that brings up the issue of trust. How much control and autonomy do you want to give an agentic AI system to make decisions on its own?โย
Based on the report, these are the four AI trends that will shape enterprises in the coming years.
- The upending of entire technology systems
The first trend relates to the technology development cycle. To fully take advantage of AI, companies will need to redesign their end-to-end systems โ a move that will fundamentally change their enterprise architecture and lay the foundations for infusing AI deeply into their organisation.
AI will not just deliver automation on top of existing business operations, but can create new processes, workflows and software. Viale pointed to how companies such as Adobe, Microsoft and Salesforce have begun to transform and enhance their software suite by equipping it with more AI capabilities. Users remain in control of the experience but are exposed to features like AI agents, chatbots and other integrated tools.
Brunier cited the example of deploying an agentic AI system at a bank to fulfil tasks that were previously performed entirely by relationship managers. The system would work independently with minimal human intervention, providing advice on product and services. Relationship managers would only be brought in when an in-person interaction is required.
- Personification for greater differentiation
High-quality AI tools no longer rest in the hands of Big Tech or large companies alone โ they have becomeย accessible and cost-effective enough for the majority of companies. With AI capabilities improving and more enterprises using the technology, it raises the question of how businesses can differentiate themselves โ especially when many AI interfaces tend to look and feel the same.
According to Accentureโs report, the answer lies in companies infusing AI with their values and personas as they build more agentic AI models. By doing so, they can power these AI agents with the full knowledge of the business, enabling leaders to avoid the risk of monotony.
Businesses can also harness AI personification to meet growing customer demand for more personalised products and services. Wellness company WeightWatchers and broadcast network SiriusXM, said Viale, are two examples of brands using AI in this way. Think AI-powered virtual health coaches that can help users hit their nutrition goals and AI-generated playlists tailored to a userโs listening persona respectively.
โThe next step would be to manage, at scale, the human dimension in terms of customer interactions with greater specificity, and for enterprises to be able to apply this to all people equally and with a personalised, personified element,โ added Viale.ย
- Robotics gets a reinvention
In the next few years, physical manifestations of AI such as robots and humanoids will become more sophisticated and play an increasingly important role in the enterprise. As Viale mentioned, companies are drawing on the growth of GenAI, 3D graphics and other supporting technologies to give robots more autonomy in the physical world, especially along the manufacturing and industrial chain. This allows them to interact with people, have a better grasp of their environment, and understand complex instructions and take safe and accurate actions in response.ย
Viale brought up the example of an automotive manufacturer that has deployed around 500 specialised robots in its manufacturing line. โCobotsโ โ collaborative robots built for direct human-robot interaction within a shared space โ are another trend heโs observed, along with โdark factoriesโ that are fully automated to facilitate 24/7 production with minimal human intervention.ย
According to Brunier, another trend to watch is the use of digital twins โ comprehensive models that enable companies to run digital simulations based on real-time conditions, which can help them make better decisions and optimise their processes.
โYouโre basically building the full manufacturing area in a digital twin, piloting and testing before you deploy anything in real life,โ he explained. โThis is one of the key investment areas and has massive disruption potential.โ
4. AI as a collaborative learning technology
Many workers worry about theย threatย AI poses to their employment. However, as the report states, AI isnโt the same as automation technologies of the past. The uniqueness of GenAI in particular is that itโs a learning technology that becomes more useful the closer it is to humans. If approached correctly, leaders can trigger a learning loop between people and AI that puts employees in the driverโs seat for innovation.
Viale emphasised that this could provide new opportunities, help organisation achieve greater human-machine complementarity and optimise organisational agility. Use cases range from deploying AI to construct self-serve workflows for onboarding data scientists, which e-commerce company Wayfair has done, or to support the development of new drugs and vaccines in healthcare.ย
To pull this off,ย tech know-how among an organisationโs leadership is essential. โWe see a major impact on the tech fluency required in the C-suite to really grasp AI reasoning and learn from the technology,โ Viale said. โThis necessitates new, different skills to really understand the human-machine interaction.โ
The challenges of AI adoption
Many companies currently lack the technical pre-requisites to adopt and scale AI effectively. Businesses should establish the right data foundation and quality while simultaneously deploying AI tools, or the ROI could be too slow for some to stomach, said Brunier. To make this as seamless as possible, the CIO and CTO should be given seats on the executive board โ as opposed to reporting to the CEO or CFO โ so they have more influence in the process.
To determine which AI technologies to adopt amid the dizzying possibilities, Viale stated that enterprises should rely on numbers and data to run a value tree analysis. As with what businesses did during the digital wave, they need to quantitatively examine the value pool to determine where AI can best generate revenue and address pain points along the value chain.
โIn three years or so, the solutions we build around AI will be set up to do human-like reasoning,โ Brunier forecasted. He added that controlling the technology and determining how much autonomy to afford it will be an important challenge for the C-suite.
Indeed, as AI becomes more powerful, maintaining human autonomy will be critical. Brunier offered his take: โAutonomy is for us to decide. It’s for us to determine the level of control we give AI and the governance we apply. This will ultimately determine how quickly and effectively enterprises are able to scale the technology.โ
Edited by:
Rachel Eva Lim
About the author(s)
is the Dean of Research and Innovation, the UBS Chaired Professor in Investment Banking and a Professor of Finance at INSEAD. In addition, she directs the
Finance for Executives course and the
INSEAD Fintech Programme.
-
Original Post
Related: You CAN Manage, Forecast, and Evaluate AI Costs
Related
