Custom AI Solutions that work: A human-centred approach

Many organisations recognise the benefits of custom AI solutions but face difficulties in their effective implementation. A recent MIT Technology Review Insights report highlights this notable disconnect. While nearly half of executives believe customised AI can help them achieve competitive advantage, an overwhelming 76% find it challenging to identify appropriate use cases. This disparity between ambition and reality prevents organisations from unlocking AI's full value.

Understanding Custom AI Solutions

Custom AI solutions are artificial intelligence systems designed to address specific organisational challenges rather than off-the-shelf products with generic capabilities. Unlike standardised AI tools, custom AI solutions aim to integrate with a company’s unique workflows, data structures, and business objectives.

These bespoke solutions typically involve adapting or fine-tuning existing AI models or building entirely new ones to meet specific requirements. They may incorporate an organisation's proprietary data, industry-specific knowledge, and particular use cases that generic AI systems cannot adequately address.

What distinguishes effective custom AI solutions is their ability to solve real business problems while seamlessly integrating into existing operations.  The MIT report highlights AT&T's implementation as a prime example. The telecommunications giant integrated custom AI solutions to help migrate legacy code which allowed them to streamline rote and repetitive tasks. This enabled employees to focus on more complex, higher-value work that directly support their mission of providing better customer connectivity and service. Other organisations have achieved similar results, with legal firms improving document analysis efficiency and marketing agencies like Dentsu reducing media analytics processing time by 90%

As the MIT Technology Review report indicates, the most successful implementations focus not just on technical capabilities but on delivering tangible value to the people who use them.

The struggle to implement Bespoke AI

While many organisations recognise the value of custom AI solutions, many struggle to implement them. Dr Hayley Watson, our Director of Innovation and Research, explains this disconnect stems from organisations placing too much emphasis on technology and neglecting the critical human element:

Successful AI customisation needs to be designed around people and workflows, not just technology. In our experience, AI projects must prioritise the user experience and real-world needs from the start. In practice, this means the AI is built to solve meaningful workplace problems in ways that feel natural for employees. By focusing on how people actually work and what they care about, organisations ensure the AI system delivers value in a way that users accept and trust. A sociotechnical approach—blending social and technical factors—makes the AI a helpful assistant rather than a disruptive novelty that will receive resistance within the workplace.

The MIT Technology Review report reinforces Dr Watson's perspective, observing a shift among forward-thinking organisations from measuring AI success solely by technical criteria to evaluating the outcomes for human users. Organisations are increasingly asking not "What can AI achieve?" but "How can AI empower people to achieve more?". The key is starting with such questions during the design phase, rather than evaluation phase.

Cross-functional Teams: A collaborative solution to implementation challenges

Companies at different stages of AI maturity struggle with different concerns when implementing their custom AI solutions. According to the MIT Technology Review report, these include data privacy (52%), poor data quality (49%), and challenges evaluating AI's real-world effectiveness (45%). Dr Watson underscores the necessity of cross-functional teamwork to effectively navigate these challenges.

Custom generative AI solutions require teamwork across many disciplines, bringing together AI engineers, software developers, data specialists, domain experts, ethicists and data protection experts. Collaboration among these specialists is essential to tailor the AI to the business, mitigate risks as it relates to bias, data protection and data security and the nuances of the domain. This cross-functional effort ensures the AI is responsibly designed, technically sound, aligned with business goals, and incorporated into existing systems. When diverse teams work in unison, the resulting AI fits the organisation's needs, meets compliance requirements and is more robust and reliable.

This collaborative approach directly addresses the issues highlighted in the MIT report. Data privacy concerns diminish significantly when privacy experts are integral to development teams. Data quality concerns lessen when domain and data specialists collaborate closely. Measuring AI's effectiveness becomes clearer when business stakeholders actively engage with technical experts in setting concrete, human-centred success criteria.

Cross-functional teamwork is essential to effectively navigate the challenges mentioned above. It brings together technical, ethical, and domain expertise to ensure AI solutions are both technically sound and properly integrated into existing business processes.

Making Custom AI Solutions work for people

The MIT report identifies a clear pattern: organisations understand the value of AI yet struggle with successful practical application. Dr Watson argues that a human-centred perspective is typically the missing link in developing custom AI solutions.

At Trilateral, we have many years of experience of The importance of embedding human-centred perspectives throughout the AI development process is clear to us. This approach allows us to produce effective solutions that integrate naturally into existing workflows. This integration is the key to delivering genuine improvements and earning user trust.

Organisations seeking to maximise ROI from their custom AI solutions should:

• Centre AI design around human needs from the outset.
• Foster cross-disciplinary collaboration to address complex implementation issues.
• Position AI as a practical tool supporting real-world business needs, not as a mere technological novelty.

By placing people at the centre, organisations can finally bridge the gap between ambition and successful implementation.

To explore human-centred AI and responsible implementation strategies further, contact us at sociotech@trilateralresearch.com or visit our Bespoke AI page.