Don’t Start with Data: 3 Tips for Implementing AI Analytics
Focusing on business outcomes, aligning teams, and achieving quick wins can help your AI analytics program succeed.
- By Pete Reilly
- November 18, 2019
Data is critical to AI analytics, but striving for data perfection on the front end can needlessly halt the implementation of AI. Companies can pour millions of dollars into data warehouses and data lakes without getting closer to actionable insights or data-driven cultures that positively impact the bottom line. In other words, it’s all too easy to get bogged down in data prep.
A strong AI analytics strategy must instead center on business outcomes. An outcomes-based approach ensures that businesses are driving AI analytics toward goals that really matter -- and that data is modeled based on these outcomes.
Before I discuss the minutiae of a successful AI analytics implementation, I want to briefly explain the value of AI analytics itself.
What is AI Analytics and Why Does it Matter?
According to AnswerRocket’s AI Analytics Guide:
AI automates the steps that humans would take to complete analysis in an exhaustive fashion. AI can test every possible data combination to determine hierarchies of relationships between different data points -- and it can do so much faster than a person could.
AI analytics accelerates data analysis and handles the bulk of the work. The output of the analysis is translated via natural language generation (NLG) so users receive insights in plain English. The combination of NLG and machine learning to automate insights is referred to as augmented analytics, which far exceeds the capabilities of traditional analytics tools and dashboards.
To put this in context, traditional data analytics largely relies on the expertise of data analysts and scientists. Traditional platforms provide data visualization tools and dashboards that organize and present data in an easily consumable form.
More advanced analytics workflows -- hypothesizing, testing data combinations, reviewing results, identifying trends, and piecing visualizations together into a complete narrative -- are performed by technical employees. Where dashboards and traditional analytics show data, AI analytics interprets data to uncover meaningful insights.
AI performs the actual analysis, determining the root causes behind metrics and surfacing hidden insights and findings. As such, AI analytics enables business people, not just technical team members, to understand their data and make insights-driven decisions.
Now, let’s talk implementation.
Tip #1: Start with business outcomes
AI analytics should center on specific business outcomes. After all, AI solves problems and identifies opportunities. To successfully implement AI analytics, businesses should think in terms of the results that the analysis can help drive. Broadly, these outcomes could be improving revenue, managing risk and cost, and/or innovating new offerings.
If improving revenue is the goal, companies need to refine their understanding of sales and marketing performance as well as the competitive landscape. The kind of analyses that would contribute to this understanding could include:
- Root cause analysis that tells employees why sales and marketing metrics are in decline or on the rise, not just what those metrics are
- Proactive analysis that monitors KPIs for meaningful fluctuations and alerts team members when something important occurs
- Analysis that pinpoints opportunities to increase revenue generation and prioritizes them based on value
Likewise, these analyses are tied to workflows at a department or employee level. Companies should consider the day-to-day actions employees could perform that would move the needle on driving top-line revenue. For example:
- The sales team needs a tool that enables them to ask data-related questions and get answers that make sense to them without the intervention of an analyst. This tool should be optimized for mobile and voice accessible so sales teams can ask questions on-the-go as they move from meeting to meeting.
- The marketing team needs fast insights to get up-to-date performance data on active campaigns so team members can pivot accordingly.
- Data and analytics employees need freedom to build custom machine-learning models instead of running routine reports for sales and marketing (just like a data scientist stuck running the same churn-prediction algorithm over and over again).
From here, it can be useful to frame these actions in terms of the decisions that need to be made and the questions that need to be answered to achieve desired results. For example, a marketer who can ask questions such as “What are inbound leads by channel in Q1?” can quickly and effectively decide where to funnel their resources for upcoming and current campaigns.
To sum up, stakeholders should answer the following questions in this order:
- What is the big picture outcome we want to achieve for the organization?
- How is this outcome supported at the functional level via cascading department objectives?
- Which key decisions need to be made and which questions need to be answered to make progress on these department objectives?
- What data is needed to inform those decisions and answer those questions?
Following this method helps ensure that the data ultimately supports the larger business outcome and enables AI analytics to maximize business impact.
Now that we’ve discussed why companies shouldn’t start with data, let’s explore the critical change-management strategies that should be used in tandem with determining the business outcome.
Tip #2: Align data and business teams on AI analytics
One of the challenges of implementing AI analytics is getting all stakeholders on board. As with any technological shift, alignment between business people and technical team members is critical to ensure long-term usage.
Interestingly, AI provides an opportunity for stronger alignment than less-advanced analytics solutions. If companies invest in analytics software that doesn’t automate the output of analysis, they may run into unexpected walls.
Furthermore, when data analysts present their findings to businesspeople, businesspeople don’t have the context of the hypotheses tested and the research performed. In this sense, the output of the data analysis may be as dense and confusing as the data itself. Receiving analysis without transparency or the ability to ask follow-up questions can leave businesspeople without clear action items.
Likewise, data analysts who are often bogged down in reporting backlogs may not have the time to fully understand each question and its context. In this case, both data analysts and businesspeople are interpreting each other’s work. With cutting-edge AI analytics, the insights produced will be consistent, contextual, exhaustive, and understandable.
Businesspeople get their questions answered quickly and in a straightforward manner. Data analysts can free themselves from repetitive reporting and leverage AI analytics for more advanced, custom machine learning models. Thus, both businesspeople and technical teams stand to gain from AI analytics, but both parties need to agree upon and champion a solution.
First, businesspeople must understand how AI analytics will help them be more effective in their roles. Automation enables business people to act quickly and focus on the creative work that’s unique to humans (and generally more enjoyable). Businesspeople become citizen data scientists who are empowered to make data-driven decisions.
On the data side, employees need to validate the AI analytics solution to ensure that it:
- Maintains the organization’s data security standards and best practices
- Offers enterprise-grade data governance functionality, such as centralized semantic models and metadata that users can leverage
- Supports AI and machine learning libraries with open extensibility
- Can scale beyond its first use case to support the larger goals of the organization and the unique needs of different departments
Data and analytics professionals need assurance that the solution will support their expertise and uphold businesswide analytics standards.
Just like businesspeople, these professionals must understand that automation is not equivalent to displacement. They, too, can become more effective in their roles and make progress on business objectives.
Tip #3: Get quick wins to maintain momentum
In What Artificial Intelligence Can and Can’t Do Right Now, Andrew Ng, a leading expert in AI, writes, “If a typical person can do a mental task with less than one second of thought, we can probably automate it using AI either now or in the near future.” These “one second tasks” are prime material for a fledgling AI analytics implementation.
A market share or brand health analysis can be broken down into a series of one-second tasks that can be improved with automation. For example, entering data into an Excel cell or adding a line of text to a PowerPoint presentation of data visualizations are both components of the larger analysis -- and the burden of these tasks multiplies exponentially with the amount of data to be analyzed.
In addition, there’s a notable difference between a stagnant visualization on a PowerPoint slide and an interactive visualization that allows the user to drill down into specific data points and ask follow-up questions with immediate answers.
Switching from Excel and PowerPoint to an AI analytics solution can drastically improve workflow. In fact, this change encompasses much more -- the ability for employees to accomplish data analysis that would take days (or weeks) in seconds and the democratization of data for nontechnical employees.
Employees who experience the speed and ease of AI analytics and offload time-consuming, repetitive tasks are more likely to advocate for broader implementation. That’s doubly true when the insights produced are highly actionable and help identify where users should focus their attention to gain the most impact.
The best AI analytics solution won’t be impactful if stakeholders aren’t on board and if employees aren’t using it. Automation that enhances workflow and helps employees be more effective in their jobs can ultimately support top-level business outcomes that improve performance.