Practice Safe AI with Domains
Generating Diplomacy
Business analysis is like a diplomatic mission, forging productive communication between stakeholders to get a solution right the first time. Generative AI can supercharge this communication by generating well-informed requirements, user stories, and other content.
Generative AI can enrich research from the industry level down to the domain level. A domain is an area of expertise within an organization, usually with specialized terms, concepts, and processes. Salesforce, for example, has cloud solutions for three domains: sales, marketing, and customer service. A domain can contain subdomains, each with more specific expertise.
Applying generative AI on a domain and subdomain basis can create higher-quality business analysis content with less risk.
Practice Generative AI to Approach Perfection
Generative AI, like ChatGPT, may seem as exciting to business analysts (BAs) as a solution that makes users’ jobs less painful and more productive. BAs may feel tempted to dive headlong into generative AI, like anxious stakeholders clamoring for a life-improving solution. They should resist this temptation and collaborate with AI to create useful deliverables such as requirements, user stories, and stakeholder interviews.
BAs must first carefully develop prompts - directives for generative AI to create content. This development process takes several test and refinement cycles to generate good enough deliverables. Cutting corners on prompt development is like giving users an untested prototype - a likely disappointing result. Diligent BAs invest time and energy to make prompts and prototypes worthwhile.
Domain-Driven AI Testing
Creating specific prompts for each domain or subdomain focuses an AI’s training, guiding it to generate high-quality content with enough testing and refinement.
AI prompts usually require examples from a domain, such as process steps, business needs, or even strategic context. When using a public generative AI like ChatGPT, BAs must ensure that proprietary information does not go into any prompts, especially information that gives the organization a competitive advantage.
Domain types can guide BAs to choose safe information to use in prompts. Domains and subdomains have three types:
A Core Domain contains strategic solutions that differentiate the organization from its competitors.
A Supporting Subdomain contains common solutions required by a core domain.
A Generic Subdomain contains standard solutions for other subdomains, like identity services or reporting.
BAs should avoid using information from the core domain in generative AI prompts. Doing so risks valuable content falling into a competitor’s hands or even training a public AI model with your organization’s competitive advantage. For example, a business analyst would not get ChatGPTs help on user stories that will give their organization a competitive edge.
Instead, BAs should use content from generic and supporting subdomains in their generative AI prompts. These domain types offer ample scope for experimentation while keeping organization secrets safe. For instance, they could use GPT’s help with authentication or standard integration requirements.
Create Valuable Content, not Risks
Business analysis provides a crucial link between business needs and IT solutions, and introducing generative AI can significantly enhance this connection. Business analysts should focus AI prompts on the lowest domain and subdomain levels to create specific, high-quality requirements, user stories, and stakeholder interviews. Additionally, BAs should use information only from generic and supporting subdomains, steering clear of core domains to protect organizational secrets and competitive advantage.
Generative AI can revolutionize business analysis with domain-based guidance to craft specific prompts while keeping confidential information safe.