‘The future is here; it’s just unevenly distributed’: New report assesses corporate AI efforts in Maine

BNN’s Business Intelligence Manager, Darren Fishell, attended the State of AI in Maine conference at the Roux Institute on January 27. The conference detailed the findings of a new report on the current state of corporate AI in Maine. Fishell, a data science master’s student at the Institute and a former business reporter, shared the following recap of the conference and his key takeaways from the report.

On January 27, hundreds of professionals from across Maine gathered to learn about and discuss how businesses are using artificial intelligence techniques to drive decision-making and efficiency in their businesses.

The promise of AI has captured imaginations in a new way, with a wave of generative AI technologies that can write, design, and make us question: what is creativity? For instance, the language model ChatGPT has given many a specific example of a new way of retrieving information, via conversation.

If you haven’t heard of this language model from the company OpenAI, you should give it a try.

It can write knowledgeable and concise essays; it can write or translate computer code; it can even try to write poetry or, when prompted just right, write fan fiction for popular TV shows or movies.

In short, ChatGPT has made a lot of the talk and promise of AI real, making a convincing case that a computer really can synthesize information and communicate in a way that resembles human thought.

While ChatGPT is a great example, companies across Maine are leveraging more established and simpler AI techniques to help make decisions and uncover insights hiding in their data, according to the Roux report. The Institute has also helped companies accelerate development and deployment of more advanced tools.

The report includes specific recommendations about deploying AI based on interviews with fifty Maine companies representing five industries: healthcare and life sciences; natural resources; education; insurance and financial services; and manufacturing.

Key considerations for a successful AI deployment

Full-scale automated decision-making takes significant testing and performance review before being given any responsibility. But fully automated decision-making is often not the right solution.
The report details how companies like Norway Savings Bank have automated rote tasks that can be expressed in the form of specific “If [situation], then [action].”

Christine McCann, assistant VP at the bank, told the researchers that this has not taken away work from employees, but “free[d] them up to focus on the things we want a trained mind to be able to focus on.”

The report identifies four recommendations for approaching and developing AI tools, which come at various levels of complexity:

  1. Lay the groundwork (data, infrastructure, expertise)
  2. Involve all stakeholders in the ‘test and learn’ process
  3. Nurture computational subject matter experts
  4. Start with a targeted use case or proof-of-concept

1. Laying the groundwork

A company’s core data systems are often siloed and not prepared for analysis. Companies first need to integrate and learn the structure of their data, reshaping it to suit the types of analysis they want to produce.

This is often an iterative process of exploratory data analysis, where continued probing of the data leads to more and more refined questions.

Laying groundwork for advanced analysis involves specific technical resources and involvement of the business to help identify specific problems that could potentially be solved with automation or AI solutions.

At BNN, we have approached this groundwork in a way that drives some immediate value, by continually integrating and enriching new data sources while building flexible business intelligence reports.

At times, this integration alone can help solve data quality issues and start identifying areas where there is not a single source of truth.

Understanding and assessing data quality is a key part of laying this groundwork. You’ve perhaps heard the data adage: garbage in, garbage out. Building visibility for data quality is another key part of laying this groundwork.

2. Involve all stakeholders

At BNN, our Technology Team has launched two new steering committees with our tax and assurance practices. These committees will meet quarterly to review, identify, and prioritize strategic information and technology initiatives, with a particular focus on process improvement.

These efforts are still in early stages, but it’s a critical part of making automation, analytics, and AI deployment a success. You need the right people at the table in conversation or you risk missing key data sources and uses, and not achieving organizational buy-in.

David Messinger, a director and actuary at FullscopeRMS, echoed this sentiment in the report when he told researchers that “it’s incredibly important that the stakeholders are not just a point of contact, but are really integrated through the whole arc of the project.”

He noted that it can be a large time commitment for teams that have plenty of day-to-day activity to focus on, but he said that involvement has been the most important component of a successful project.

3. Nurture computational subject matter experts

The Business Intelligence group at BNN is a relatively new part of the business, and we’re seeing this reflected across companies in Maine and beyond. More and more, leaders see their operational data as an asset.

The establishment of a Business Intelligence group within BNN was a first step toward process automation and artificial intelligence, but it’s critical to blend those skills with an understanding of our business problems and business context.

This takes stakeholder involvement to its logical conclusion: rather than having IT serve discrete reports, enable subject matter experts to work directly with validated data models.

Many BI tools, like PowerBI, now make this possible in the tool of a user’s choosing. Enabling self-service reports on high-quality and validated data sources is a great way to reduce reporting bottlenecks and to involve more of the business in the exploratory analysis that leads to better questions from subject-matter experts.

How this works and who is involved will be different at every organization. It’s important to take time to assess who your subject matter experts, stakeholders, and technology and data specialists are, and make sure you are assessing automation opportunities with the full picture of what the business needs.

4. Start with a targeted use case or POC

Businesses can’t rush to AI adoption—success here takes time. In my experience and hearing from others at the conference, a “crawl, walk, run” approach to the development of BI tools is the best way to achieve success when it comes to new technology, tools, and processes.

Replicating existing reports is a great place to start for these kinds of projects, especially where the data is extracted or created manually and then transformed by hand.

This is the first area where automation can be helpful. Report-building in modern BI tools can automate these steps: extracting, transforming, and presenting the data in a consistent format and process. Beyond that, they give the end-user levers to influence those steps, such as filtering a report by department or drilling down to a specific period.

These can be great starting points that can also be extended into action. For instance, a newly automated data pipeline can provide alerts to specific users based on specific thresholds.

Where do we go from here?

Even if AI is not your organization’s end goal, the journey towards an AI deployment itself holds numerous valuable lessons and insights that are worth pursuing in their own right. You will learn things and discover elements of your operations, employees, and processes along the way. And you’ll be ready for what’s next. Involve the right people and don’t rush the process. The businesses that are most successful with automation and data are approaching this topic with curiosity and an open mind.

If you’re interested in hearing more detailed examples, I encourage you to read the full State of AI in Maine report here.

Disclaimer of Liability: This publication is intended to provide general information to our clients and friends. It does not constitute accounting, tax, investment, or legal advice; nor is it intended to convey a thorough treatment of the subject matter.