Having spent my 25+ year career in digital and technology marketing, I’ve seen my share of initiatives not meet expectations. AI initiatives in marketing, be they Generative, Predictive, or other type of AI technology are posing a new set of challenges and raising the stakes.
There's no doubt, there are aspects of AI that are unique, and this article attempts to focus on those. But issues common to AI project shortcomings are going to sound very familiar to those in technology-oriented functions. The issues typically include at least one element related to people, process, technology, and/or data not being adequate, and generally all situations include a breakdown in communication at some point.
Let’s look at a few issues that can trip up AI initiatives. This is not an “agile how-to” or “better project management” post, but it does touch on those. Rather, I’ll present takeaways, tips, and recommendations to address issues to correct or minimize a project from going sideways.
Issue: Misunderstanding Data Models and Data Lifecycles?
- Develop educational programs and resources that explain AI data models and life cycles. These assets should be tailored for all levels involved (i.e. beginner, functional lead, C-Level). Improving AI and data literacy across the organization is an investment that will pay off.
- Improve documentation and visibility of the data lifecycle within AI powered products and systems. Honest documentation builds trust and enables better decision-making among product teams.
- Encourage cross-functional collaboration between data scientists, AI SMEs, and business stakeholders to ensure alignment and a common language of AI system needs and business objectives. Organization supported Communities of Interest or AI workgroups can be helpful.
- Adopt a hands-on approach to learning. Involve team members in the data curation and model refinement processes. Emphasize the incremental and iterative nature of AI solutions.
- Leverage Agile methodologies to guide AI development. Seek leadership support for an environment where learning about data models and lifecycles is part of the improvement and review cycle.
Issue: Treating AI Projects Like Software Development
- AI really needs a dynamic and iterative approach that includes change management, with emphasis on data management for decision-making and adaptation based on real-time insights.
- Successful AI integration into an organization's operations tends to be better with an ecosystem perspective. One that ensures new tools or assets align with downstream and existing processes and infrastructures.
- Unlike past approaches to software products or projects, AI systems need ongoing maintenance and tuning. The occasional patch or app update isn’t sufficient. Rather a cultural shift towards continuous learning and improvement within teams is needed.
- The coming impact of AI on job roles and work processes means organizations need strategies that prioritize upskilling and create pathways for an evolving workforce. Don’t hide from the fact that an AI system will have an impact on how some jobs are done. Be upfront with those who may be impacted.
- The Agile methodology, with its focus on collaboration, iteration, and efficiency, is particularly relevant for managing the fluid nature of AI project development and integration. Adopt Agile elements that align with organizational standards or project needs.
Issue: Data Management for Quality & Quantity
- Prioritize data governance and infrastructure to maintain a strong foundation for data collection, cleaning, and management. Without this the quality and usability of data for AI projects is compromised.
- Build a data-centric culture with the organization. Doing so builds on good data governance. Employees who understand the value of data assets are more likely to contribute high-quality data.
- Embrace continuous improvement as an ongoing process which follows the guidelines established with a strong data governance foundation. (Strong = focused + collaborative)
- Balance quantity and quality of data. Initial project success maintains its importance, but future scalability can be addressed to some extent within the framework of data governance and continuous improvement.
Issue: Challenges Managing Marketing Data for AI
- Develop a strategic data management plan that includes data acquisition, storage, processing, security, and governance, tailored to AI's needs within a marketing context (i.e. consumer data).
- Invest in data infrastructure that ensures data is accessible, secure, and compliant with relevant privacy regulations (i.e. HIPPA, CCPA, etc.), to enable proper AI system(s) functionality.
- Leverage data management platforms that support the integration and normalization of data, providing clean, organized datasets for AI applications.
- Develop a culture of data literacy within the organization to ensure all stakeholders understand the importance of high-quality data for AI success.
Issue: It Worked in R&D but Not in Production
- Allow for a more agile R&D process. Include frequent checkpoints and adjustments based on feedback from environments that simulate actual production conditions.
- Support improved communication between stakeholders with cross-functional collaboration. Require R&D, operations, and product teams to work closely together to identify and address potential issues before full deployment.
- Prioritize transparency and communication between teams to ensure that everyone has a realistic understanding of the capabilities and limitations of the AI being developed.
- Establish continuous integration/continuous delivery (CI/CD) pipelines for AI, enabling regular and automated testing of AI models in production-like environments.
- Create a feedback loop with real users early in the R&D phase. Iterate and monitor the AI models based on insights gathered to ensure they are adequate and work as intended in real-world scenarios.
Issue: Pilot Projects Take Too Long to Realize Potential
- Set Specific, Measurable, Achievable, Relevant, and Time-bound (SMART) goals for pilot projects. Set clear metrics that can be evaluated frequently and with ease.
- Use data from similar past projects both within and outside the organization. Use these to inform expectations and identify potential roadblocks early.
- Apply an iterative approach, breaking down the project into smaller, testable segments. Gather insights and refine predictions as the project progresses. Do this on a regular schedule.
- Involve stakeholders from various departments to provide diverse perspectives and insights, which can help in creating more realistic projections. Users and Builders will have differing opinions about success.
- Regularly review and adjust projections based on actual performance and feedback during the pilot phase. Don’t wait until the end to evaluate success. No one likes surprises after being told everything has been fine.
Issue: Benefits from AI are Over Promised and Under Delivered
- Over optimism in AI's capabilities can lead to a disconnect between the promises made by vendors or development teams and the actual utility and performance of AI tools in the wild.
- The complexity and newness of AI applications require significant resources and can lead to gaps between expected and delivered benefits. This can be particularly true when AI is applied to sophisticated projects like personalization in healthcare or finance.
- AI's developmental lifecycle is continuous and requires persistent tuning and learning. The requirements can be underestimated in the face of market pressures to deliver quick results, not to mention hype surrounding AI that adds to pressure and expectations.
- Enthusiasm and excitement of AI can result in overlooking the nuances in implementation, including data quality, model robustness, and the alignment of AI objectives with overall business strategy.
- The hype generated by media coverage can overshadow the methodical, iterative, and complex processes required for successful AI projects. This leads to inflated expectations and subsequent disappointment in the short term if not met. Over communicate as is necessary.
Issue: Buying Into Vendor Hype
- Conduct comprehensive proof-of-concept tests that align with specific organizational needs. Leverage R&D teams that aren’t end buyers/users who can independently assess the actual effectiveness of AI products and systems beyond the hype.
- Engage with community forums, industry groups, and/or networks to gain insights and performance benchmarks from peers who have implemented similar AI solutions. Every organization is in an AI-learning phase right now.
- Use an Agile approach to implementation, starting small and scaling up only after verifying the initiative delivers tangible benefits in its real-world application.
- Develop internal competencies in AI and data science to critically evaluate vendor claims. Verify the potential and limitations of AI technologies. This is a great team activity to collectively build greater understanding by sharing findings across internal groups.
- Focus on measurable outcomes and ROI from AI investments. Adjust expectations and strategies based on actual outcomes and experiences. Try to tune out unverified industry buzz.
Issue: Return on Investment (ROI) Misalignment
- Ensure AI initiatives are directly linked to solving key business problems or exploiting specific opportunities to drive measurable outcomes.
- Confirm that AI is the right tool for the task at hand. Assess its fit based on the strategic goals and current capabilities of the organization.
- Quantify the impact by developing a set of metrics for measurable ROI that are grounded in the reality of the company's operations and the AI's potential performance.
- Rank the potential ROI metrics by their relevance to decision-makers and by the initiative’s ability to help achieve the organization’s long-term vision. Speak truth to power if these are out of alignment.
- Present AI not only as a path to financial returns but also as an investment in the organization's future capabilities and maintaining a competitive edge.
Outlook for AI Initiatives in Marketing
Can you feel it? It’s not a literal earthquake but it is a seismic shift. While it is still early days for adoption and operationalizing AI, broad market momentum is picking up. The First Movers have already left the starting gate and the Early Adopters are honing capabilities. Yet the vast majority are still figuring out what AI potentially means to their organizations. The promise AI holds is not just in its current capabilities but in its potential to evolve with us, reshaping how we understand, engage, serve with our audiences. By proactively addressing the challenges common in early-stage technology adoption, the humans who’ll carry out these initiatives will be in position to effectively leverage the possibilities AI can offer. For marketing organizations this means delivering more value to the right audience at the right time for the right level of investment.
Resources for Further Investigation
#AI #ArtificialIntelligence #MachineLearning #MarketingAI #AImarketing #MarketingTechnology #DataManagement #DataGovernance #DataQuality #ProjectManagement #AgileMarketing #ChangeManagement #AIethics #ResponsibleAI #AItransparency
