Key ingredients companies need to successfully integrate AI

August 12, 2020

Blog Insights Trustworthy AI

Photo by Scott Graham on Unsplash

August 12, 2020

Unlike startups and tech giants with AI-first strategies, most companies don’t have AI labs and teams singularly dedicated to finding AI application opportunities and adapting research for production.

For such companies still keen to gain the benefits of AI adoption, a prerequisite for success is well-developed AI receptor capacity. AI receptor capacity is a term Vector’s Industry Innovation Team uses to denote the expertise and approach required to effectively integrate AI into an organization’s existing operations.

One essential feature of AI receptor capacity is alignment among important AI project stakeholders within an organization. Promising AI activities often require attention at several levels of management within a company, and each may have its own priorities. Those working with the code may have questions about the Python package needed for model training. Managers guiding them may be deciding whether to further curate data or dedicate a team to developing a specific kind of model. Executives overseeing these managers may be focused on prioritizing AI goals according to competitive dynamics within the industry.

Clear communication of each stakeholders’ priorities is crucial to giving AI experimentation the best chance at success. However, there can be challenges here. Executives will likely find it difficult to discuss AI goals with technical teams in terms of hyperparameter selection and other technical concepts. Conversely, technical teams may find it a challenge to explain how broad strategic goals for the company must be refined and framed as specific problems suitable for AI.

It’s crucial to have a person or team that can manage this alignment and communication. Such a person must be capable of walking seamlessly between the technical and business domains within the company. A name sometimes used to describe people in this role is business translator.

Effective business translators have certain credentials and characteristics in common, the first of which can be found on their CVs. They typically have degrees and experience in technical fields, including computer science, engineering, math, or physics. Their experience and skills should also demonstrate a history of being close to the code and a familiarity with important AI frameworks for developing models, such as Tensor Flow and PyTorch. Ideally, they keep current with academic publications and trends in the field and can understand how those trends impact the sector.

Part of their role is also helping organizations to set up processes to avoid the innovator’s dilemma. AI experiments can be inadvertently crushed by the regular processes, politics, and risk management approaches of the company. AI initiatives must be guarded from such a fate. When there isn’t a separate lab insulating initiatives from such pressures, it’s often up to business translators to help protect them and make sure they get the resources, attention, and space they need.

 Beyond facilitating alignment and protecting early adoption efforts, business translators can help companies evaluate vendors soliciting AI services. Fast-moving, newly-commercialized technologies are often accompanied by a flood of vendors of varying quality. Services that shoehorn business problems into inappropriate models, repackage open source code as commercial solutions, or purport to solve complex AI issues like explainability or fairness require scrutiny by skeptical and informed professionals. Business translators can reduce information asymmetry between the company and vendors, and help ensure procurement decisions are driven by informed analysis.

AI is impacting every industry. Companies — even in industries not historically considered first adopters of new technology — need to keep track of adoption among industry peers and seek out models, services, and talent accordingly. However, awareness and intention while adopting is not sufficient for progress. In order to turn adoption into value, organizations need to go beyond interest and set the stage to thrive. A well-developed AI receptor capacity, anchored by business translators, will help turn AI into value for the firm.


Vector Faculty Member Gautam Kamath

Vector researcher Gautam Kamath breaks down the latest developments in robustness and privacy

Trustworthy AI

World-leading AI Trust and Safety Experts Publish Major Paper on Managing AI Risks in the journal Science

Large Language Models

Standardized protocols are key to the responsible deployment of language models