top of page

Planning for Success with Generative AI

12/24/23

Editorial team at Bits with Brains

As generative artificial intelligence becomes increasingly practical to implement, careful consideration is key to avoid potential pitfalls and realize true value.

As we all know by now, while the potential upside is large, generative models also pose novel risks that demand proactive planning from leadership.


A recent Forbes survey found generative AI topping boards' priority lists, with 46% labeling it their highest concern more than anything else. But before declaring it a top initiative, decision-makers must critically assess alignment with strategic goals and allocate a proportionate level of resources. A KPMG study revealed most adopters anticipating returns within 3-5 years invest significantly upfront.


With limited resources, choosing the right starting data is vital. The survey highlights customer, market research and labeled datasets as having the highest comfort levels among respondents. Beginning with high-quality information lays the best foundation for useful, trustworthy results.


It's also critical to define user groups and the training needed to ensure proper handling. As one leader framed it - providing the tools isn't enough without education on their responsible operation. Surveys highlight designating a single point of responsibility as a leading implementation strategy, whether that role falls to the CEO, CDO, or CTO.


Accompanying technology access, initiative-taking policy creation around privacy, fairness and transparency forms another key tenet of governance. Regular audits, clear accountability and staff education on AI ethics help uphold standards as use broadens across teams.


Communicating objectives and results transparently also factors prominently in building understanding and trust. By answering these fundamental questions up front, leaders can pave the way for generative AI to reach its potential while avoiding pitfalls through prudent oversight and planning.


Here are the top questions to consider before implementing Generative AI.

  1. What is generative AI's priority compared to other initiatives? Once determined, ensure you allocate appropriate resources.

  2. How much should you invest? It is best to plan on a 3–5-year ROI.

  3. What data do you have that you can use? It is best to select high-quality data sources like customer, market research and other labeled data.

  4. Who will be the Generative AI users? Start by defining the intended user groups, as the power must be coupled with proper training.

  5. Who will manage implementation? Designate a singular implementation leader – domain experts are best.

  6. What policies need to be created? Who will be responsible for proactively creating policies around privacy/security, fairness/bias, and trust/transparency?

  7. Who will enforce governance? Enforce governance through regular audits, clear responsibility lines and staff training on AI and AI ethics.

  8. How do you plan to educate employees about GenAI usage? Provide education tailored to user groups on risks, transparency, and hallucination prevention.

  9. What are the expected benefits? Outline benefits like time savings and increased sales to measure against.

  10. How do you plan to communicate the results? Commit to communicating results transparently across the organization.

Here are several related considerations for leaders when planning for generative AI implementation.


First, data selection is paramount. Choosing data sets directly tied to key business metrics allows organizations to quantify generative models' impact. Starting with transactional or customer satisfaction data, for example, facilitates measuring outcomes like sales lift or customer experience improvement.


Next, clearly defining success criteria upfront also helps ensure generative projects deliver tangible benefits. Goals like time savings, cost reductions or revenue growth should be specific, measurable, and relevant to the use case. This allows for objective evaluation of models' performance over time.


Third, pilot programs offer a controlled testing ground before full deployment. Beginning with a limited scope and user group provides learnings to refine processes, address issues and prove value before widespread adoption. Pilots can inform further investment and expansion plans.


Fourth, multi-phased rollout strategies stage the introduction of more advanced capabilities. For instance, conversational AI functions may precede creative generation or process automation in complexity. Gradual, gated adoption based on pilot results mitigates risks from rushing full deployment.


Fifth, careful change management is essential as generative technologies disrupt existing roles and workflows. Close collaboration between leadership, IT and affected teams to redefine job duties and upskill workforces facilitates smooth integration of new systems and tools.


By addressing data, success metrics, pilot testing, change management and rollouts in a phased, systematic way, organizations can deploy generative AI solutions confidently and capture their true potential for transformation.


As in M&A, prudent planning paves the way.


Sources: 

https://www.forbes.com/sites/forbestechcouncil/2023/12/20/10-questions-to-ask-before-implementing-generative-ai/

Sources

bottom of page