What AI Success Looks Like

Most enterprise AI initiatives stall after the pilot phase.

One of the world’s most esteemed financial organizations, Goldman Sachs, took a different route.

In mid-2024, the firm launched the GS AI Assistant—a generative AI platform designed to emulate the behavior of seasoned employees.

Built on models like OpenAI's ChatGPT and Google's Gemini, the assistant was layered with proprietary compliance, security, and training infrastructure. It wasn’t a tool to experiment with—it was designed to be used.

The assistant was rolled out to 10,000 employees across operations, engineering, and client-facing teams. It wasn’t dropped in from above. It was seeded into workflows that already existed. Teams used it to draft investment briefs, summarize research, prep client updates, and write or translate code.

The result: faster output, fewer bottlenecks, and more strategic use of human time. Within six months, the firm saw double-digit efficiency gains in multiple departments.

This is what AI success looks like. And here’s why it worked.

☕️ AI Tangle - Meta's First AI Developer Conference, Google's AI Mode Rollout & Visa's Push For Shopping Via Chatbots

🔮 AI Lesson - Prompting for Reasoning Models

🎯 The AI Marketing Advantage - Why Your LLM Won’t Push Back

💡 AI CIO - What the Latest Verizon DBIR Says About AI

🎙️ AI Confidential Podcast - Building Transparent, Open-Source AI with Sriram Raghavan

 📚 AIOS - This is an evolving project. I started with a 14-day free Al email course to get smart on Al. But the next evolution will be a ChatGPT Super-user Course and a course on How to Build Al Agents.

What AI Success Looks Like

Most AI projects stall because of poor implementation. Here's what tangible success looks like—and what your organization needs to achieve it now.

Not every company is a Wall Street institution. But the principles behind their success apply universally—especially now that AI isn't a fringe experiment. It's a basic expectation.

Tobi Lütke, CEO of Shopify, captured this shift bluntly. In a company-wide note, he told staff: “I consider it the responsibility of every knowledge worker in Shopify to understand how to use AI and to figure out how to integrate it into their work.” This wasn’t an optional initiative. It was a cultural mandate. And it signals where things are headed.

If your org isn’t building toward that expectation, it’s already behind.

88% of AI Pilots Fail

That’s the current state of enterprise AI. According to a report authored by IDC on behalf of Lenovo, nearly nine out of ten AI projects never make it past the pilot stage. The reasons are familiar: no executive ownership, low user adoption, weak infrastructure, and use cases that don’t solve real problems.

Goldman flipped the script by choosing use cases that were:

• High frequency

• High friction

• Measurable

• Non-customer-facing

The assistant wasn’t launched to replace jobs or transform strategy. It was designed to augment what employees were already doing—repetitive document generation, internal communication, and routine analysis.

It also wasn’t imposed from the outside. The tool was purpose-built, with guardrails for regulated workflows. It had access to the right data. Its outputs were traceable and compliant.

By anchoring the initiative in reality, the firm avoided the most common trap: building AI without a business case.

Start Where Value Is Obvious

Generative AI tools excel at summarization, drafting, and translation. Goldman leaned into this by identifying workflows that were already document-heavy, error-prone, or slow to scale.

One example is internal investment memos. Before the assistant, analysts might spend hours compiling data, writing context, and formatting output. With the assistant, the base document could be generated in minutes. Analysts then edited and refined—acting as reviewers, not just authors.

The same applies to software development. Engineers used the tool to translate legacy code, document APIs, and respond to support tickets. Time spent on rote communication dropped. Output quality increased.

The impact wasn’t just faster work. It was more meaningful work. Employees could spend more time on the hard parts—decision-making, relationship-building, and strategic planning.

Buy-In Wasn’t an Afterthought

One reason AI efforts stall is the lack of top-down alignment. Not at Goldman. CEO David Solomon championed the initiative from the beginning, positioning AI as a workforce enhancer—not a threat.

The same top-down mandate is now emerging elsewhere. Shopify’s Tobi Lütke isn’t just encouraging AI adoption; he’s demanding it. He’s said that the company itself is becoming a “digital by default, AI-native” organization. That kind of framing shifts the burden from IT to every department. AI isn’t a tool to test—it's a tool to master.

This approach works because it communicates two things clearly:

(1) AI is here to stay, and (2) your career trajectory depends on learning how to use it.

The Right Use Case Doesn’t Need a Press Release

Successful companies aren’t chasing headlines. They’re fixing internal friction. Goldman didn’t launch an AI chatbot for customers or a branded microsite. They targeted boring but valuable tasks.

Shopify did the same. They started with internal tools like Sidekick—a merchant-facing AI assistant—and followed with deeper workflow automation across their engineering and operations stack. No hype cycles. Just applied intelligence.

The lesson is simple: optimize for use, not attention.

AI Is Only Useful If People Use It

A technically powerful tool that no one touches is worthless. Usability is the differentiator.

Goldman made the assistant available inside tools employees already used. It didn’t require learning a new UI. It spoke plain English. It surfaced relevant content from proprietary systems. And it respected internal rules of engagement.

Shopify’s directive mirrors this approach. Tools are embedded in Slack, GitHub, documentation workflows, and even onboarding. The principle is: bring AI to the work, not the other way around.

Upskilling Is Essential

Most companies drop AI tools on teams without teaching them how to use them. According to Accenture’s report, Making Reinvention Real with Gen AI, organizations spend three times more on technology than on people.

Companies investing heavily in AI tools without training their employees is like giving a state-of-the-art surgical robot to a hospital but without training any surgeons to operate it. This results in a multi-million-dollar machine sitting unused in the corner while patients continue to receive conventional treatment—or worse, doctors attempting to use it through trial and error at patients' risk.

The result: low adoption, poor prompts, and inaccurate or fabricated output ("hallucinations").

The firms getting it right are investing in structured, role-specific training. Goldman gave engineers, analysts, and team leads tailored materials and examples. Shopify is doing the same—with Lütke calling AI fluency the next "Excel for the modern worker."

This kind of enablement builds a culture of practical experimentation, not AI tourism.

Success Compounds When AI Fits The Org

AI doesn’t succeed because it’s novel. It succeeds because it aligns with how a company already works.

Goldman’s assistant was built to match the firm’s norms—precision, traceability, high-quality output. It didn’t try to reinvent workflows. It aimed to accelerate them.

Similarly, Shopify’s success depends on embedding AI into its product, engineering, and customer functions—not adding layers, but removing friction.

That’s the model: not reinvention, but acceleration.

Takeaways for enterprise AI teams

If you're planning a similar AI rollout, start here:

1. Pick a narrow, painful use case. Look for tasks that are repetitive, measurable, and non-customer-facing.

2. Get executive buy-in early. AI strategy without leadership support is just a suggestion.

3. Build guardrails from the start. Model choice matters less than data access, prompt engineering, and compliance.

4. Integrate into existing systems. Reduce friction. Make the AI show up where users already work.

5. Upskill by role. Don’t throw a prompt guide over the wall. Train people based on how they work.

6. Track and communicate success. Share wins, highlight outputs, and keep adoption visible.

You don’t need to boil the ocean. You need to solve something real.

Goldman did. That’s what AI success looks like.

Many organizations have already invested in platforms like Microsoft 365 and Google Workspace. These suites now come equipped with AI capabilities that can enhance productivity without the need for additional tools. While I still think that OpenAI is the best option, there’s inherent value in using tools that are already part of your workflows. I also added Fyxer because I am piloting it right now. I think the ability to take notes and to draft personalized email responses is essential for improving productivity and streamlining communication across teams.

• Microsoft 365 Copilot - Microsoft 365 Copilot integrates AI into familiar applications like Word, Excel, PowerPoint, Outlook, and Teams. It assists with drafting emails, summarizing meetings, generating documents, and analyzing data. By leveraging large language models and Microsoft Graph, Copilot provides contextual assistance based on your data and interactions.​

• Gemini for Google Workspace - Gemini integrates AI-powered assistance throughout Google Workspace applications including Gmail, Docs, Sheets, and Slides. This intelligent assistant enhances productivity by supporting content creation, analyzing data, and generating meeting summaries. By offering real-time suggestions and handling repetitive tasks automatically, Gemini streamlines workflows across the Workspace environment. The system leverages generative AI technology to boost collaboration efficiency throughout the entire suite of Google productivity tools.

• Fyxer - Fyxer AI acts as an AI executive assistant that organizes your inbox, drafts personalized email responses, and generates meeting notes. It learns from your communication style to provide tailored assistance, integrating with Gmail and Outlook. Fyxer AI saves time by automating email management and meeting documentation, allowing teams to focus on higher-value tasks.​

Collaborative Goal Setting with ChatGPT

When you're navigating high-stakes decisions—whether it's launching a new course, boosting revenue, or positioning for a promotion—what you need most is a clear-thinking, unbiased collaborator. Large language models can be exactly that: a structured partner for strategic dialogue, not just a content generator.

Introduction

Most professionals don’t need another brainstorm. They need a business-savvy counterpart to test assumptions, uncover blind spots, and help drive toward measurable outcomes. That’s where generative AI—when used correctly—can function as a neutral strategy assistant.

Whether your goal is launching a product, optimizing operations, or advancing your career, LLMs can help you define the problem, explore constraints, and develop a tactical plan. The key is treating the AI less like a search engine and more like a business partner—one who always answers the phone, never takes credit, and thinks in structured logic.

In this article, we’ll walk through a repeatable method to use AI for business problem-solving. You’ll learn how to define your objective, capture key constraints, and translate your goals into an actionable plan—powered by AI prompts that drive results.

Here's an interactive prompt that will help you brainstorm and drive success in your business endeavors. Just cut and paste it into your favorite chatbot and start iterating.

## 🧠 AI Use Case Definition Interview

### 👤 Role  
You are a **Business Strategy Assistant** that helps professionals define AI use cases and generate powerful AI prompts to meet business goals.

---

### 🎯 Your Objective  
- Interview the user to clarify:
  - **Business objective**
  - **Current state and blockers**
  - **Target audience** and **measurable outcome**
  - **Resources and constraints**
  - **Appropriate AI approach** (e.g., GenAI, RAG, automation)

- **Confirm each input** with the user before continuing.

- After the user confirms “**done**,” output:
  1. A tailored **AI Prompt** based on their use case.
  2. A detailed **Execution Plan** broken into four implementation phases:
     - **Align and Prioritize**
     - **Data and Tools Prep**
     - **Prompt Engineering and Iteration**
     - **Deployment and Training**

I appreciate your support.