Blogs

Navigating the AI Advisory Storm: Persistence, Persuasion, and Practical Solutions

As an AI advisor navigating the turbulent waters of innovation, each day brings a new set of challenges and opportunities. Let me take you through a typical day in this exhilarating, often demanding role.

The Daily Grind: A Balancing Act

My day often starts with a list of open items - unresolved questions, pending decisions, and looming deadlines. It's a mix that requires both technical depth and strategic thinking. Here's how it usually unfolds:

Morning Research: I dedicate time to reading the latest AI research and industry trends. In this field, staying current isn't just beneficial - it's essential.
Experimentation: Hands-on work with new models or frameworks keeps my advice grounded in practical reality.
Solution Architecture: The bulk of my day involves designing solutions that bridge ambitious visions with technical feasibility.

The Art of Persistence and Persuasion

In the world of AI advisory, two skills stand out as crucial:

Persistence: The path from concept to implementation is rarely smooth. It requires tenacity to push through setbacks and keep projects on track.
Persuasion: A significant part of my role involves convincing stakeholders to make tough decisions. This might mean persuading a founder to adjust their vision based on technical limitations, or convincing a development team to push harder to meet an important deadline.

Navigating Tricky Waters

Some of the most challenging aspects of this role involve:

Design Trade-offs: Every decision has consequences. I often find myself weighing the pros and cons of different approaches, always mindful of both short-term needs and long-term scalability.
Bugs vs. Deadlines: The eternal struggle in tech takes on new dimensions in AI projects. Balancing the need for robust, bug-free systems against the pressure of tight deadlines requires careful judgment and clear communication.
Bridging Diverse Perspectives: In any given day, I might jump from a high-level strategy discussion with a visionary founder to a nitty-gritty technical debate with the development team. The ability to speak both languages and translate between them is crucial.

The Reward in the Challenge

Make no mistake - this role is demanding. The constant pressure to innovate, the need to make high-stakes decisions with imperfect information, and the rapid pace of change in the AI field can be overwhelming.

But it's also incredibly rewarding. There's a unique thrill in being at the forefront of technology, in seeing ideas transform into solutions that can change industries or even societies.

A Call to Action

For those considering a role in AI advisory, or looking to work with advisors in this space, here's my advice:

Cultivate a learning mindset. The field changes too rapidly to ever rest on your laurels.
Develop your communication skills. The ability to explain complex concepts simply is invaluable.
Practice resilience. Not every idea will work out, but each setback is a learning opportunity.
Stay ethical. With great power comes great responsibility, and AI is undoubtedly a powerful tool.

The world of AI advisory is not for the faint of heart. It's a storm of innovation, challenges, and opportunities. But for those willing to weather it, the view from the eye of the hurricane is unlike any other.

My Perspectives - GenAI Adoption Advisory: Key Principles and Strategies

1. First Principles Thinking: Connect Basics to Innovate, Especially Under Time Constraints

In my GenAI advisory roles across various domains (Industrial, Real Estate, Fashion), I've learned that each project requires a fresh perspective. My approach involves:

Structuring problems uniquely for each client
Questioning assumptions about existing processes and technologies
Brainstorming innovative options by connecting fundamental concepts
Building tailored solutions from the ground up
Innovating creatively when time is a constraint

Example: For Proplens AI, I reimagined the entire real estate workflow for sales, marketing, and customer service. By connecting basics of AI, CRM, and real estate processes, we created an innovative solution using Azure ecosystem, Claude Sonnet 3.5, GPT-4, and Neo4j Graph, all within tight timelines.

2. Hunger for Impact: Leverage Past Experiments for Quick, Tangible Results

As an AI advisor, I'm driven by the desire to create substantial, measurable impact:

Develop end-to-end AI engines that solve real business problems
Bridge the gap between cutting-edge AI technologies and practical business applications
Mentor teams to build AI competency from the ground up
Apply lessons from past experiments to rapidly derive tangible products

Example: For the Industrial Data Management AI project, I leveraged insights from previous data catalog projects to swiftly develop an intelligent system using LLMs and custom NER models. This approach allowed us to quickly scale the solution across supply chain, procurement, and data management sectors.

3. Depth of Analysis: Prototype and Test Reasonably

While maintaining a high-level strategic view, I ensure thorough analysis through:

Conducting in-depth market analysis to identify key opportunities and challenges
Performing detailed technical evaluations of AI models and architectures
Preparing comprehensive appendices for deep-dive discussions
Creating prototypes to test ideas reasonably and efficiently

Example: For the Vision + Fashion AI startup (Fitumi), I conducted extensive analysis and rapid prototyping of various AI models (Stability AI, OpenAI, Gemini). This approach allowed us to quickly identify the most effective solutions to minimize sampling, marketing, and product development costs.

4. Brief and Direct Communication: Provide Perspective with Deep Understanding

When advising executives, I focus on:

Providing synthesized insights with clear, measurable impact
Highlighting key decision points and strategic implications
Preparing concise summaries with the option to delve deeper if requested
Offering perspective backed by deep understanding and relevant examples

Example: In my AI Advisory role for AuditOne GmbH, I deliver focused recommendations on model monitoring, content moderation, and compliance. By providing real-world examples and demonstrating a deep understanding of the challenges, I help executives make informed decisions quickly.

5. Centering on Strategy: Look Ahead and Leverage Domain Knowledge

I always frame AI adoption in the context of broader business strategy by:

Aligning AI initiatives with overall business objectives
Emphasizing the "why" behind AI adoption and its potential impact
Regularly revisiting and refining the strategic vision as the project progresses
Looking ahead rather than just comparing current products
Using domain expertise as a springboard for next-level innovations

Example: For the Meal Waste Reduction project with Unilever, I centered the GenAI solution development around the core strategic goal of sustainability. By leveraging domain knowledge in FMCG and looking ahead to future sustainability trends, we ensured that the AI implementation not only contributed to immediate waste reduction targets but also positioned Unilever as a leader in sustainable food management.

By adhering to these principles, I've successfully guided multiple organizations through their GenAI adoption journey, from initial strategy development to full-scale implementation and optimization. My approach combines innovative thinking, rapid prototyping, and strategic foresight to deliver tangible results in the fast-paced world of AI.

16 Changes to the Way Enterprises Are Building and Buying Generative AIIn 2024, enterprise leaders are doubling down on their genAI investments. 16 developments for founders to keep in mind to capture this new opportunity.

GenAI + Vision

Plant Identification Using Convolution Neural Network and Vision Transformer-Based ModelsIdentification of plants is a challenging task which aims to identify the family, genus, and species level according to morphological features. Automated deep learning-based computer vision algorithms are widely used for identifying plants and can help users to narrow down the possibilities. However, numerous morphological similarities between and within species make the classification difficult. In this paper, we tested a custom convolution neural network (CNN) and vision transformer (ViT) based models using the PyTorch framework to classify plants. We used a large dataset of 88K and 16K images for classifying plants at genus and species levels respectively. Our results show that for classifying plants at the genus level, ViT models perform better compared to CNN-based models ResNet50 and ResNet-RS-420, and other state-of-the-art CNN-based models suggested in previous studies on a similar dataset. The ViT model achieved top accuracy of 83.3% for classifying plants at the genus level. ViT models also perform better for classifying plants at the species level compared to CNN-based models ResNet50 and ResNet-RS-420, with a top accuracy of 92.5%. We show that the correct set of augmentation techniques plays an important role in classification success.