aiSTROM -- A roadmap for developing a successful AI strategy (2107.06071v2)
Abstract: A total of 34% of AI research and development projects fails or are abandoned, according to a recent survey by Rackspace Technology of 1,870 companies. We propose a new strategic framework, aiSTROM, that empowers managers to create a successful AI strategy based on a thorough literature review. This provides a unique and integrated approach that guides managers and lead developers through the various challenges in the implementation process. In the aiSTROM framework, we start by identifying the top n potential projects (typically 3-5). For each of those, seven areas of focus are thoroughly analysed. These areas include creating a data strategy that takes into account unique cross-departmental machine learning data requirements, security, and legal requirements. aiSTROM then guides managers to think about how to put together an interdisciplinary AI implementation team given the scarcity of AI talent. Once an AI team strategy has been established, it needs to be positioned within the organization, either cross-departmental or as a separate division. Other considerations include AI as a service (AIaas), or outsourcing development. Looking at new technologies, we have to consider challenges such as bias, legality of black-box-models, and keeping humans in the loop. Next, like any project, we need value-based key performance indicators (KPIs) to track and validate the progress. Depending on the company's risk-strategy, a SWOT analysis (strengths, weaknesses, opportunities, and threats) can help further classify the shortlisted projects. Finally, we should make sure that our strategy includes continuous education of employees to enable a culture of adoption. This unique and comprehensive framework offers a valuable, literature supported, tool for managers and lead developers.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Explain it Like I'm 14
What is this paper about?
This paper introduces aiSTROM, a clear, step‑by‑step roadmap that helps companies plan and run successful AI projects. It exists because many AI efforts fail or get abandoned, often due to weak planning. aiSTROM shows managers and tech leads what to think about—from data and team setup to ethics and success measures—so they can pick the right projects and make them work.
What questions does it try to answer?
In simple terms, the paper asks:
- How do we choose the best AI projects to start with?
- What data do we need, and how do we keep it safe and legal?
- Who should be on the AI team, and where should that team sit in the company?
- Should we build AI ourselves or use ready‑made services?
- How do we deal with risks like bias and “black‑box” models that are hard to explain?
- How do we measure progress and value (not just profit)?
- How do we build a culture where people actually use and trust AI?
How did the author approach it?
This is a practical guide, not a lab experiment. The roadmap is based on:
- The author’s decade of leading AI projects
- A scan of useful research and industry reports
Think of aiSTROM like a recipe for a big group project at school: you pick the right tasks, gather the right materials, assign roles, choose tools, check your progress, and keep everyone learning.
The aiSTROM roadmap in plain words
The paper suggests first listing several possible AI projects, then taking the top 3–5 and checking each one across key areas:
1) Identify opportunities
- Pick problem ideas that matter (save time, cut costs, improve products, or create new ones).
- Choose a few with big payoff and reasonable effort.
2) Data strategy
- Do we have the right data? If not, can we collect, buy, or label it?
- Understand “big data” basics: lots of data (volume), arriving fast (velocity), from many places (variety), with quality concerns (veracity), and real value (value).
- Handle privacy and security (follow laws like GDPR, CCPA, PDPA); reduce breach risks.
- Decide how and where to store data (e.g., data lake vs. data warehouse; cloud vs. in‑house).
3) Build the AI team
- Mix skills: math/stats, coding, data engineering, design, communication, and constant learning.
- Hiring is hard; consider up‑skilling current staff, partnering with universities, or “acquihiring” (buying a small company mainly for its team).
4) Organize development
- Place the team in the company: centralized (one hub), decentralized (each department has its own), or hybrid “hub‑and‑spoke.”
- Use a “portfolio” of projects (some low‑risk quick wins, some longer‑term bets).
- Decide build vs. buy: use AI‑as‑a‑Service (AIaaS) when it’s not core to your business.
- Use agile methods and MLOps: build small prototypes, test with users, improve quickly.
5) Technology choices
- Accuracy vs. explainability: some models are very accurate but hard to explain (“black‑box”). Sometimes the law or fairness demands explainable models.
- Keep humans in the loop: people can label data, review results, and guide the system.
- Replace vs. augment: aim to help people do their jobs better, not just replace them.
- Compute power: rent cloud GPUs to start fast, or buy hardware later if it pays off; consider small AI chips for devices (IoT) when needed.
6) KPIs (success measures)
- Define simple, meaningful scorecards tied to your goals (e.g., happier customers, faster service, fewer errors), not just money.
- Track model accuracy too (e.g., how often it’s right), and be ready to adjust if results are weak.
7) Risk and readiness
- Rate each project’s strengths, weaknesses, opportunities, and threats (SWOT).
- Match choices to your company’s appetite for risk.
- Plan for ongoing training so people adopt and trust AI.
What are the main takeaways and why do they matter?
- Start with the right projects: Picking a few high‑impact, realistic ideas saves time and money.
- Data first: Good, legal, secure data is the fuel of AI; without it, models fail.
- Team and structure matter: A skilled, mixed team placed well in the company gets things done faster.
- Build or buy wisely: Don’t reinvent the wheel—use AIaaS when it’s not your core advantage.
- Be ethical and clear: Avoid biased results, follow privacy laws, and use explainable models when needed.
- Measure real value: Track benefits customers feel and the efficiency you gain, not only revenue.
- Learn continuously: People need training and time to accept and use AI; that’s how AI actually delivers value.
These points matter because many AI projects fail from poor planning, not poor tech. Following aiSTROM raises the odds that AI helps the business and its customers in a safe, fair, and useful way.
What could this change in the real world?
If companies follow aiSTROM:
- More AI projects will succeed, bringing better products and smoother services.
- Workers can do more meaningful tasks with AI as a helper, not a replacement.
- Customers get fairer, clearer, and more private experiences.
- Companies stay competitive in a fast‑moving economy.
- Organizations build a long‑term culture of learning and responsible innovation.
In short, aiSTROM is a practical, people‑focused plan for making AI work well in real companies, today and over time.
Collections
Sign up for free to add this paper to one or more collections.