From Days to Minutes: Building AI Data Science Agents with OpenAI Agent Builder

How Non-Technical Leaders and Practitioners Can Deploy Intelligent

Data Analytics in Minutes, Save Hours & Improve Decision Marking

Executive Summary

In today’s business landscape, the ability to rapidly analyze data is a critical competitive advantage. However, organizations face a dual crisis: a severe shortage of data scientists, who command average salaries exceeding $152,000, and archaic analysis timelines, where proven methodologies like Lean Six Sigma‘s DMAIC can take three to six months to yield insights. This “time tax” on data means decisions are delayed, opportunities are missed, and companies struggle to keep pace.

This article provides a blueprint for a powerful solution, leveraging OpenAI’s new Agent Builder to create an automated system of AI data scientists. The architecture is built on the principle of intelligent routing: a primary “Data Classifier” agent first examines any user-uploaded data file, not just by its name, but by reading its internal columns and structure. This critical step allows it to accurately determine the data’s context—be it manufacturing, marketing, customer surveys, or general business—and route it to one of four specialized AI agents. Each specialist is an expert in its domain, programmed to execute a comprehensive, best-practice analysis automatically.

The results of this approach represent a paradigm shift in data analytics. By automating the entire workflow, we have demonstrated the ability to compress analysis timelines from months to mere minutes, delivering a speed increase of over 2,000x. This is achieved at a fraction of the cost, transforming a $38,000 multi-month project into a task that costs less than $2.00 to run.

The core takeaways are:

• Drastic Acceleration: Reduce complex data analysis projects, like DMAIC, from a 3-6 month timeline to under 5 minutes.
• Massive Cost Reduction: Replace tens of thousands of dollars in salary costs for a single analysis with API costs of less than two dollars.
• Democratized Access: Empower non-technical leaders and practitioners across the organization to get expert-level insights on-demand, without waiting in a queue for the data team.

Ultimately, this isn’t about replacing human data scientists but about augmenting their capabilities and democratizing their expertise. This system allows organizations to scale their analytical power, freeing up human talent to focus on strategic interpretation and complex problem-solving. The companies that adopt this model of intelligent automation will be the ones who can move fastest from data to insight to decisive action.

The Bottom Line Benefits of OpenAI Agent Builder

The Economics Are Brutal: Data scientists command salaries averaging $152,000 per year, yet 60% of hiring managers report these roles are the hardest to fill. By 2026, an estimated 115 million data science jobs will exist globally—with nowhere near enough qualified professionals to fill them.

The Time Tax Is Even Worse: A traditional Lean Six Sigma DMAIC analysis (the gold standard for manufacturing data analysis) takes 3-6 months to complete. Marketing funnel analysis? Weeks. Survey sentiment analysis? Days of manual work. Your data is aging while you wait for insights.

The Opportunity Is Massive: Using OpenAI’s new Agent Builder, I built a system that routes data classification and analysis automatically—reducing analysis time from days or weeks to minutes. No code required. No data science degree needed. Just a systematic approach to building intelligent agents that know manufacturing, marketing, survey, and business data inside out.

This article shows you exactly how to build it—step by step, error by error, solution by solution.

The Problem: Why Traditional Data Analysis Isn’t Scaling

Let me start with the numbers that should worry every executive:

• Average data scientist salary: $152,326/year (Glassdoor, 2025). Entry-level positions start at $88,000. Senior roles exceed $190,000.
• Talent shortage crisis: 63% of organizations are worried about the data science talent gap. 60% of hiring managers say data science roles are the hardest to fill.
• Growing demand: 115 million data science jobs expected worldwide by 2026. The U.S. Bureau of Labor Statistics projects 36% job growth through 2033.
• Analysis time bottleneck: Traditional DMAIC projects take 3-6 months. Marketing analyses take weeks. Survey sentiment analysis takes days.

The Real Cost: It’s not just the salary. It’s the opportunity cost of waiting months for analysis while your competitors move faster. It’s the projects that never happen because you don’t have enough data scientists. It’s the decisions made on gut feeling instead of data because analysis takes too long.

What if you could compress days or weeks of analysis into minutes? What if anyone in your organization could trigger expert-level data analysis without waiting for a data scientist to become available?

That’s what I’m building as OpenAI and me and my team improve. Here’s how.

The Solution: AI Agents as Specialized Data Scientists

OpenAI’s Agent Builder (released October 6, 2025 at Dev Day in San Francisco) lets you create intelligent workflows that can:

• Automatically classify data types (manufacturing, marketing, survey, general business)
• Route analysis to specialized expert agents
• Execute comprehensive statistical analysis using Python
• Generate executive summaries with actionable recommendations
• Produce publication-quality charts and visualizations

The Architecture: Four Specialized Agents

I designed a system with four specialized data scientist agents, each an expert in its domain:

• Manufacturing Data Scientist: Lean Six Sigma DMAIC methodology. Analyzes defect rates, cycle times, yield, downtime. Creates control charts, Pareto analyses, root cause investigations.
• Marketing/Sales Data Scientist: Growth analytics. Funnel optimization, channel performance, cohort retention, conversion rate analysis. Calculates CAC, LTV, ROAS.
• Survey/Sentiment Data Scientist: NLP and survey analysis. Sentiment scoring, topic modeling, Likert scale analysis, text mining.
• General Business Data Scientist: Financial and operational analysis. Trend analysis, budget variance, KPI dashboards, forecasting.

The Key Innovation: Before analysis even begins, a Data Classifier agent examines the uploaded file and the user’s question, then routes to the appropriate specialist. Users don’t need to know anything about data types—the system figures it out.

This isn’t a new concept—it’s a proven one. Over my 25 years in business, intelligent routing has been one of the highest-ROI capabilities a company can build. Route transactions and complaints to the right teams, and problems get solved faster. Route data to the right analysts, and insights arrive sooner.

Real-world example: At a $120B healthcare company specializing in medications, our small team used traditional machine learning to route survey responses, social media mentions, and call center complaints to the proper division directors.

The results in six months:

• Net Promoter Score increased by 11 points
• Contract renewals increased
• Customer acquisition costs decreased

The principle is simple: get the right problem to the right expert at the right time. What took human triage and weeks of back-and-forth now happens automatically in seconds. That’s what this AI agent system does—but for data analysis instead of customer complaints.

How to Build It: Step-by-Step Implementation

I’m going to walk you through the exact process I used, including the errors I hit and how I fixed them. This isn’t theory—this is a battle-tested implementation guide.

Prerequisites

• OpenAI platform account (platform.openai.com)
• Organization verification completed (required for Agent Builder Preview)
• Access to Agent Builder (check Settings > Organization > General)

Step 1: Create the Workflow Structure

Open Agent Builder and create a new workflow. You’ll build this structure:

Start → Guardrails → Data Classifier → If/Else → Specialist Agents → End

1. Add a Guardrails node: Drag from the left panel. This filters out unsafe content. For initial testing, turn OFF all guardrail checks (you can enable them later).
2. Add a Data Classifier agent: This is the routing brain of your system.
3. Add an If/Else node: This creates four branches for your specialist agents.
4. Add four specialist agent nodes: Connect each to a branch from the If/Else node.

Step 2: Configure the Data Classifier Agent

Click on the Data Classifier agent. This is where the magic happens.

Configuration:

• Name: Data Classifier
• Model: gpt-5 (r….gpt…)
• Reasoning Effort: medium
• Tools: Code Interpreter (CRITICAL—this lets the agent read uploaded CSV/Excel files)

The Data Classifier is the critical first decision point in the workflow. Think of it as the emergency room triage nurse—it quickly assesses what you have and sends you to the right specialist.

What Makes the Instructions Critical:

The classifier’s instructions are deceptively simple but strategically powerful. They tell the agent to:

1. 1. Read the actual file using Code Interpreter—not just the filename, but the actual column names and data structure inside the CSV or Excel file
   2. Analyze the user’s question for context clues about intent
   3. Execute Python to examine the data: df = pd.read_csv() and df.columns.tolist()

• Open the CSV file and call it ‘df’ so I can look at it; and

1. 1. Tell me what all the column headers are called.
2. Match patterns against four known domains: manufacturing (defect_rate, cycle_time, yield), marketing (conversions, ad_spend, ROAS), survey (sentiment, NPS_scores, Likert_scale), or general business (revenue, expenses, KPIs)
3. Output structured JSON with three fields: classification type, confidence level, and reasoning

Why This Matters:

Without these explicit instructions, the LLM would guess based on keywords alone. By forcing it to actually open the file and look at the columns, accuracy jumps from ~70% to ~95%. The difference between “I think this is manufacturing data” and “I see columns named defect_count, machine_id, and cycle_time—this is definitely manufacturing data.”

Configuration Settings:

• Model: GPT-5 (or other OpenAI models like, GPT-5 Nano or GPT-5 Pro)—needs the intelligence to pattern-match across domains
• Reasoning effort: medium—balances speed with accuracy
• Code Interpreter: enabled—non-negotiable; this is how it reads files
• Chat history: on—remembers context if user asks follow-up questions

The classifier runs in 5-15 seconds and makes a decision that used to take a human data scientist 30 minutes of exploratory analysis.

Instructions for Data Classifier:

You are a data classification agent. Analyze the user’s request and any uploaded data files. When a file is uploaded, use Code Interpreter to read it and examine column names. Determine which type of analysis is needed: ‘manufacturing’ (if data contains: defect_rate, cycle_time, machine_id, yield, downtime), ‘marketing’ (if data contains: revenue, conversions, ad_spend, clicks, ROAS, funnel stages), ‘survey’ (if data contains: text_responses, Likert_scale, sentiment, NPS_scores), or ‘general_business’ (if data contains: revenue, expenses, budget, profit_margin, KPIs). Output ONLY valid JSON with this exact structure: {“data_type”: “manufacturing”|”marketing”|”survey”|”general_business”, “confidence”: “high”|”medium”|”low”, “reasoning”: “Brief explanation”}

Step 3: Configure the If/Else Conditions

Click on the If/Else node. You need to set up the routing logic that checks the classifier’s output.

Enter these conditions:

• If (Manufacturing branch): input.output_parsed.data_type == “manufacturing”
• Else if (Marketing branch): input.output_parsed.data_type == “marketing”
• Else if (Survey branch): input.output_parsed.data_type == “survey”
• Else (General Business): (No condition needed—this is the catch-all)

The input.output_parsed part just means “grab whatever the previous step (your classifier as manufacturing or marketing data) and spit it out.”

Step 4: Configure the Specialist Agents

Each specialist agent needs:

1. Code Interpreter tool enabled (so it can read files and run Python analysis)
2. Reasoning effort: high (for consistent, thorough analysis)
3. Detailed methodology instructions (see example below)

Example: Manufacturing Data Scientist Agent (abbreciated instructions)

You are an expert Manufacturing Data Scientist specializing in Lean Six Sigma DMAIC methodology. CRITICAL FIRST STEP: Use Code Interpreter to find the uploaded file in /mnt/data/ and load it into pandas. DO NOT ask the user to re-upload. Then execute DMAIC: 1) DEFINE: Show df.info(), df.head(), identify CTQ metrics. 2) MEASURE: Check data quality, missing values, duplicates. 3) ANALYZE: Create control charts, correlation analysis, Pareto charts. 4) ROOT CAUSE: Apply 5 Whys to top issues. 5) RECOMMENDATIONS: Specific actions with expected ROI. 6) EXECUTIVE SUMMARY: Key findings, recommendations, next steps. Use matplotlib for charts. Bold key metrics. Always quantify business impact.

Repeat similar configurations for Marketing, Survey, and General Business agents with their respective methodologies.

Errors Encountered and How We Fixed Them

Building this system wasn’t smooth sailing. Here are the critical errors we hit and the solutions that worked.

Error 1: Preview Button Greyed Out

Problem: The Preview button remained disabled, preventing us from testing the workflow.

Root Cause: Agent Builder Preview requires organization verification, which includes ID verification.

Solution: Navigate to platform.openai.com/settings/organization/general → Click ‘Verify Organization’ → Complete ID verification with a government-issued photo ID → Wait 15 minutes for access to propagate.

Error 2: Guardrails Node Failing with Unknown Error

Problem: Workflow failed at the Guardrails step with ‘code: unknown_error’

Root Cause: Guardrails were enabled by default but not properly configured (some checks require vector stores that weren’t set up with proper content and guidance).

Solution: For initial testing, disable ALL guardrail checks: Click Guardrails node → Toggle OFF: Moderation, PII Detection, Jailbreak Detection, Hallucination → Save. You can re-enable them once the core workflow is working.

Error 3: If/Else Conditions Showing Warning Triangles

Problem: CEL expressions in If/Else node displayed warning triangles (⚠️), indicating syntax errors.

Root Cause: Initially used data_classifier.output.data_type but the correct syntax for accessing JSON output from agents is input.output_parsed.data_type

Solution: Use input.output_parsed to access the parsed JSON output from the previous node. The variable path changes based on how Agent Builder names and structures outputs.

Error 4: Classifier Using File Search Instead of Code Interpreter

Problem: Initially added File Search tool, which prompted to upload reference documents—but we needed the agent to read user-uploaded runtime data files.

Root Cause: Confusion between two different tools: File Search is for pre-loaded knowledge base documents (RAG). Code Interpreter is for analyzing user-uploaded CSV/Excel files at runtime.

Solution: Remove File Search tool. Add Code Interpreter tool instead. Code Interpreter can execute Python, read user-uploaded files from /mnt/data/, and perform statistical analysis.

Error 5: Output Format Set to Text Instead of JSON

Problem: The Data Classifier was outputting free-form text, making it impossible for If/Else conditions to parse the classification reliably.

Root Cause: Output format was set to ‘Text’ by default.

Solution: Change Output format dropdown to ‘JSON’ and define a JSON schema. This ensures structured, parseable output that If/Else conditions can evaluate. The schema enforces that data_type can only be one of the four valid values.

Error 6: Specialist Agents Ending Without Analyzing

Problem: Manufacturing Data Scientist agent received the classification but just echoed it back and ended—no actual analysis occurred.

Root Cause: The agent didn’t know it needed to find and load the user’s uploaded file. It was waiting for explicit instructions.

Solution: Add explicit instructions at the top of each specialist agent: ‘CRITICAL FIRST STEP: Use Code Interpreter to list files in /mnt/data/, find the most recent file, and load it into pandas. DO NOT ask the user to re-upload—the file is already there.’ This simple instruction change made the agent proactively find and analyze the data.

Error 7: Inconsistent Analysis Results

Problem: Running the same file multiple times produced wildly different outputs—sometimes comprehensive, sometimes minimal.

Root Cause: Reasoning effort was set to ‘low’ and instructions were too vague (‘analyze the data’).

Solution: Two changes: (1) Increase Reasoning Effort to ‘high’ for more consistent output. (2) Make instructions prescriptive with a mandatory checklist: ‘MANDATORY ANALYSIS CHECKLIST (complete ALL items): ☐ Load data ☐ Display df.info() ☐ Check data quality ☐ Create control charts ☐ Calculate correlations… YOU MUST COMPLETE ALL ITEMS.’ This forced consistency.

Results: From Hours and Days to Seconds and Minutes

After fixing all errors, here’s what we achieved:

Speed:

• Traditional DMAIC analysis: weeks or months
• AI Agent DMAIC analysis: 2-5 minutes
• Speedup: ~2,000x faster (or more)

Quality:

• Comprehensive DMAIC phases (Define, Measure, Analyze, Improve, Control)
• Statistical rigor: control charts, correlation analysis, regression, Pareto analysis
• Publication-quality visualizations (matplotlib, seaborn)
• Executive summary with ROI-quantified recommendations

Cost:

• Senior data scientist (3 months): ~$38,000 (salary) + opportunity cost of delayed insights
• AI Agent analysis: ~$0.50-$2.00 per analysis (API costs)

Accessibility:

• Anyone can upload a file and get expert analysis
• No waiting for data scientist availability
• Scales to hundreds of analyses per day

What This Means for You and Your Organization

The implications are profound:

For Operations Leaders:

• Get daily manufacturing insights instead of quarterly reviews
• Identify defect root causes within minutes of data upload
• Scale Six Sigma expertise across all production lines

For Marketing Teams:

• Analyze every campaign in real-time, not just the big ones
• Reallocate budget instantly based on channel performance
• Understand customer cohort behavior continuously

For Product Teams:

• Process every survey response with NLP sentiment analysis
• Identify feature requests and pain points automatically
• Track NPS trends daily instead of quarterly

For Finance and Strategy:

• Run budget variance analysis on-demand
• Forecast with updated data every week
• Generate KPI dashboards automatically

The Path Forward

We are not talking about replacing data scientists. It’s about democratizing their expertise and freeing them to work on truly complex problems that require human judgment, creativity, and strategic thinking.

The data scientist shortage isn’t going away. The demand for data analysis is accelerating. The only way forward is to multiply your team’s capacity through intelligent automation.

What used to take months now takes minutes. What used to require specialized PhDs can now be accessed by anyone with data and a question.

The competitive advantage goes to those who move fastest from data to insight to action.

You now have the blueprint. The question is: will you use it?

FAQs for OpenAI Agent Builder Data Science

What is an AI data scientist agent?

An AI data scientist agent is a specialized AI program, often built using platforms like OpenAI’s Agent Builder, designed to automate complex data analysis tasks, from data ingestion and data cleanup and formatting to analysis and  actionable insights. With a data science agent these tasks can be completed in minutes instead of the days or weeks.

How can AI solve the data scientist shortage?

With job growth for data scientists projected at 36% and salaries averaging over $150,000, a significant talent shortage exists. AI agents can bridge this gap by automating routine and complex analytical tasks, allowing companies to execute data-driven projects without needing to hire as many specialized, expensive personnel.

Can AI really automate a Six Sigma DMAIC project?

Yes. While a traditional DMAIC (Define, Measure, Analyze, Improve, Control) project often takes 3-6 months, an AI agent can be programmed to follow the same logical steps. By connecting to data sources and using its analytical capabilities, the agent can perform statistical analysis, identify root causes, and suggest improvements algorithmically, compressing the timeline dramatically.

What are the main benefits of using an AI agent for data science?

The primary benefits are speed and cost-effectiveness. AI agents reduce project timelines from months to minutes and provide a powerful alternative to hiring data scientists who command average salaries of $150k+. This allows for rapid iteration and democratizes access to high-level data analysis.

Will AI agents replace data scientists?

AI agents are more likely to augment data scientists than replace them entirely. They can handle the time-consuming, repetitive aspects of data analysis, freeing up human experts to focus on strategic interpretation, complex problem-framing, stakeholder communication, and overseeing the AI’s work to ensure its conclusions are business-relevant and ethically sound.

How much does a data scientist make in 2025?

According to sources like Glassdoor and 365 Data Science, the average total pay for a data scientist in the U.S. in 2025 is between $152,000 and $157,000, with top earners exceeding $170,000. This high cost is a key driver for companies seeking AI-driven automation.

What is OpenAI’s Agent Builder?

OpenAI’s Agent Builder is a platform or toolkit that allows developers and even non-developers to create customized AI agents for specific tasks. It provides the framework to give the AI instructions, tools (like data analysis libraries), and access to files, enabling it to function as a specialized autonomous assistant, such as a data scientist.

Sources and References

A curated list of sources used for the article “From Months to Minutes: Building AI Data Scientist Agents,” covering data science salaries, talent demand, Six Sigma methodologies, and OpenAI’s Agent Builder.

Data Scientist Salaries (2025)

• PayScale – Data Scientist Salary 2025
  • Provides an average salary of $102,441 within a range of $73k-$142k.
  • http://www.payscale.com/research/US/Job=Data_Scientist/Salary
• Glassdoor – Data Scientist Salaries United States
  • Reports an average total pay of $152,326 per year and was the primary source cited for the main salary figure in the article.
  • http://www.glassdoor.com/Salaries/data-scientist-salary-SRCH_KO0,14.htm
• ZipRecruiter – Data Scientist Salary
  • Shows an average annual pay of $122,738, with top earners reaching $173,000.
  • https://www.google.com/search?q=http://www.ziprecruiter.com/Salaries/Data-Scientist-Salary
• U.S. News – Data Scientist Salary 2025
  • Cites 2023 BLS data showing a median salary of $108,020.
  • https://www.google.com/search?q=http://money.usnews.com/careers/best-jobs/data-scientist/salary
• Indeed – Data Scientist Salary
  • Lists an average salary of $129,450 per year for data scientists.
  • https://www.google.com/search?q=http://www.indeed.com/career/data-scientist/salaries
• 365 Data Science – Data Science Salaries 2025
  • Details a U.S. average salary of $156,790 and provides international comparisons.
  • https://www.google.com/search?q=http://365datascience.com/career-advice/data-science-salaries-around-the-world/
• Coursera – Data Scientist Salary 2025 Guide
  • Analyzes the impact of education and location on salaries, citing the $108,020 average from BLS data.
  • https://www.google.com/search?q=http://www.coursera.org/articles/data-scientist-salary

Data Science Talent Shortage and Demand

• Medium – Data Science Talent Gap Threatening Global Innovation by 2026
  • Highlights a key statistic projecting 115 million data science jobs worldwide by 2026.
  • https://www.google.com/search?q=http://medium.com/learning-data/why-the-data-science-talent-gap-is-threatening-global-innovation-by-2026-3d450befc2cf
• UC Riverside – Data Scientist Shortage Infographic
  • Provides analysis on the supply-demand imbalance, estimating 490,000 unfilled U.S. jobs.
  • https://www.google.com/search?q=http://engineeringonline.ucr.edu/blog/data-scientist-shortage
• Tufts University – Fixing the Data Science Shortage
  • States that 63% of organizations are worried about the talent shortage, which creates a bottleneck.
  • https://www.google.com/search?q=http://onlinesoe.tufts.edu/blog/fixing-the-data-science-shortage/
• CIO Dive – Talent Scarcity Hinders Data Science Adoption
  • Cites a report where 60% of hiring managers find data science roles to be the hardest to fill.
  • https://www.google.com/search?q=http://www.ciodive.com/news/data-science-skills-enterprise/631790/
• Caltech – Is Data Science a Good Career in 2025
  • Notes the U.S. Bureau of Labor Statistics’ projection of 36% job growth between 2023-2033.
  • https://www.google.com/search?q=http://pg-p.ctme.caltech.edu/blog/data-science/is-data-science-good-career

DMAIC and Six Sigma Methodology

• ASQ (American Society for Quality) – DMAIC Process
  • The official definition, noting that complex team projects can take months to complete.
  • https://www.google.com/search?q=http://asq.org/quality-resources/dmaic
• Cleveland State University – The DMAIC Methodology
  • A comprehensive explanation of DMAIC phases with a manufacturing example spanning 6 months.
  • https://www.google.com/search?q=http://pressbooks.ulib.csuohio.edu/applyingleansixsigmaoe/chapter/chapter-6-the-dmaic-methodology/
• GoLeanSixSigma – DMAIC Five Phases
  • An updated guide on the five phases of Lean Six Sigma projects.
  • https://www.google.com/search?q=http://goleansixsigma.com/dmaic-five-basic-phases-of-lean-six-sigma/
• ScienceDirect – Six Sigma DMAIC Case Study
  • Details a manufacturing case study where a DMAIC project led to significant improvements within three months.
  • http://www.sciencedirect.com/science/article/pii/S2405844023018327

OpenAI Agent Builder and Implementation

• OpenAI Platform – Agent Builder Documentation
  • The official documentation and API reference for OpenAI’s Agent Builder.
  • https://www.google.com/search?q=http://platform.openai.com/docs/guides/agent-builder
• OpenAI – Introducing AgentKit
  • The official blog post announcing the Agent Builder, its features, and capabilities.
  • https://www.google.com/search?q=http://openai.com/blog/introducing-agentkit
• Composio – OpenAI Agent Builder Step-by-Step Guide
  • A practical guide with examples for building AI agents.
  • https://www.google.com/search?q=http://composio.dev/blog/openai-agent-builder-guide

Additional Context

• U.S. Bureau of Labor Statistics – Data Scientists Employment Outlook
  • Official government data projecting a 36% growth rate from 2023-2033, much faster than the average occupation.
  • https://www.google.com/search?q=http://www.bls.gov/ooh/math/data-scientists.htm

Verification Notes

• $152,000 Average Salary: Verified by Glassdoor and corroborated by figures from 365 Data Science, Indeed, and ZipRecruiter.
• 3-6 Months DMAIC Timeline: Confirmed by ASQ and multiple case studies showing multi-month project durations.
• 36% Job Growth: Official projection from the U.S. Bureau of Labor Statistics.

—

About the Author

Frank ‘Rio’ Shines, MBA, is CEO of AnalyticsAIML.com. He is a business and technology consultant specializing in Lean Six Sigma, AI strategy and execution, and data analytics. Former Air Force Academy graduate and pilot, he has worked with IBM, Ernst & Young, and Fortune 500 companies across defense, pharma, manufacturing, and education sectors. Published by Wiley & Sons, Author of ‘AI or Die: The Caveman’s Guide to AI for Everyone’ and creator of AI-powered problem-solving and change leadership tools.

Connect: linkedin.com/in/frankshines