Challenge
Our client, a pioneering manufacturing company, has developed a groundbreaking procedure for metal processing that combines the precision of laser technology with the productivity of traditional punching. Despite its innovative approach, the technology remains largely unrecognized within the industry, leading potential customers to opt for lower quality methods and providers. This obscurity poses a significant challenge for the client, hindering their ability to attract customers and achieve growth. As a medium-sized entity in the industrial sector, expanding the sales and marketing teams to bridge this awareness gap is not a viable option due to resource constraints. Consequently, the client is determined to shorten sales cycles by employing artificial intelligence to precisely identify potential customers based on the distinct product and production characteristics that align with their unique technology. The ambition is to harness an advanced AI research agent to augment the efforts of the sales team, operating under the hypothesis that such a tailored AI solution can vastly outperform traditional human-led research efforts, significantly improving efficiency and effectiveness in identifying compatible business opportunities.
Solution
To address the challenge of connecting our client's innovative metal processing technology with the right customers, we developed a semi-autonomous AI research agent. This sophisticated solution is engineered to sift through thousands of patents, employing advanced analytical capabilities to understand and reason over the extensive data. It is specifically designed to draw comparisons between the benefits and features of our client's technology and the information gleaned from patents. By maximizing an objective function, the AI research agent efficiently identifies potential customers who are most likely to benefit from the unique manufacturing process.
Moreover, the agent extends its search beyond patents, conducting advanced web searches to cover a multitude of data sources, thereby ensuring a comprehensive analysis of the market. The application of advanced machine learning techniques allows the AI to distill the vast amounts of collected unstructured data into actionable insights. This process culminates in the preparation of a comprehensive research paper that is both easy to understand and instrumental for the sales team to plan their next steps. Through this approach, the sales team is equipped with a powerful tool that not only identifies potential leads but also provides them with a deep understanding of each lead's needs and how our client's technology can meet those needs, streamlining the sales process and significantly shortening sales cycles.
AI Tech Stack
- Finetuned GPT-3.5 for better agentuc reasoning
- Open-source embedding models from huggingface
- FAISS in-memory vectorstore
- Gradio frontend
- Our custom Advanced web search component compatible with many state-of-the-art LLMs
Our Approach
In tackling the challenge presented by our client's unique position in the metal processing industry, our approach diverged from the conventional path of seeking new business opportunities. Instead, we focused on optimizing an existing internal process with the primary objectives of enhancing efficiency and reducing operational costs. This strategic choice was driven by the recognition that significant improvements in these areas could directly contribute to the client's growth and competitiveness in the market.
To ensure the success of this project, we placed importance on the seamless integration of our solutions into the client's existing workflows. It was essential that the AI-driven enhancements could either effortlessly complement the current processes or establish entirely new, more efficient workflows. This dual focus on compatibility and improvement required a meticulous analysis of the client's operational methodologies, enabling us to design solutions that were not only innovative but also practical and immediately beneficial.
The solution's fit within the broader ecosystem of the client's operations was carefully considered throughout the development process. Our team worked closely with the client to ensure that the implementation of the AI research agent would not conflict with existing operations but instead would improve them, making the transition as smooth as possible. By prioritizing efficiency, cost reduction, and seamless integration, we aimed to deliver a solution that was not just a technological advancement but a practical, scalable, and sustainable enhancement to the client's operating model.
DISCOVER
In the discovery phase, our approach was characterized by a concentrated focus on a smaller, internal target group rather than casting a wide net over the external market. This allowed for a more detailed and intimate understanding of the specific needs and challenges faced by the employees directly interacting with our technology.
Acknowledging the critical role of user experience and journey, we thoroughly examined these aspects to ensure that any proposed solution would meet the sales teams' expectations and requirements. This examination involved a detailed decomposition of the current process to identify its strengths and weaknesses, providing a clear picture of areas ripe for improvement.
To further refine our understanding and approach, we tested various features and unique selling points (USPs) of the technology. This iterative process of testing and analysis enabled us to gather valuable insights into what modifications would be most beneficial. Through this methodical approach, we aimed to lay a solid foundation for developing a solution that not only addresses the identified challenges but also aligns with the users' needs and preferences.
PROTOTYPE
During the prototyping phase, we prioritized the integration of a proof of concept and a feasibility study at the earliest stages. This strategic approach was adopted to mitigate risks and avoid the incurrence of unnecessary costs should the main technical hypothesis not hold true. By validating our assumptions early, we were able to proceed with confidence or pivot as needed without significant resource investment.
A blueprint of the target process was crafted, serving as a roadmap for the development and testing phases. This blueprint not only outlined the process at hand but also detailed the interactions between different components and the expected outcomes, providing a clear vision for all team members.
Before committing to the full-scale development of the new process, we conducted manual tests to ensure its feasibility. This step was crucial for identifying potential issues and refining the process without the complexity and expense of software development. Through a cycle of building, testing, and iterating, we gradually enhanced the process, continually aiming for the expected sales research agent. This iterative approach ensured that when we were ready to build and scale, the foundation was solid, significantly increasing the likelihood of successful implementation and adoption.
CRAFT
In the crafting phase, we assembled a team consisting of three key roles: an AI engineer, a full-stack developer, and a business architect. This lean team structure was designed to ensure agility and efficiency, enabling us to navigate the development process effectively.
The craft phase was structured around five sprints, each lasting two weeks. This sprint-based approach allowed us to maintain a steady pace of progress, with defined goals and deliverables for each period. It facilitated a focused and disciplined workflow, with the flexibility to adapt as insights were gained and challenges encountered.
Regular stakeholder meetings were a critical component, ensuring alignment and transparency throughout the project. These meetings provided opportunities for feedback, allowing us to refine our approach and ensure that the evolving solution remained closely aligned with the client's needs and expectations.
Intermediate checkpoints for testing were strategically placed throughout the development timeline. These checkpoints served to evaluate the solution's performance and functionality against predefined criteria, allowing for timely adjustments and enhancements. This iterative and cyclical methodology contributed to a successful outcome for the project.
