Written by Will Rush, Co-Founder of Breakwater
About two years ago, I was visiting my dad, Don Rush, a fourth-generation ship supplier at Delaware Ship Supply. I had spent the previous four years building a venture-backed AI education company that partnered with more than 600 colleges and universities across North America to bring AI into business education. The experience had taught me a great deal about AI before ChatGPT became a household name, but I was now looking for my next challenge.
That was when something struck me.
The industry I had spent the previous 15 years working in – the U.S. technology sector – was completely redesigning their operating model to stay ahead of the largest technology shift in a generation. Leading tech companies were rethinking org structure, workflows and the very definition of human work in an AI-driven future.
Meanwhile, my dad’s 120-year-old family business remained focused on serving ships through time-tested workflows that had not changed in decades.
The need for change was not obvious. Business was good. The company had an exceptional reputation, a loyal team and a healthy P&L. But the way quotes were completed, orders were accepted and items were sourced was largely unchanged. Key employees had been with the company for decades. And despite using a well-known ERP system, there were artifacts of an aging process: calculators, stacks of paper and a magnifying glass out on someone’s desk.
I grew curious about bringing what I knew from one world into the other. I also knew I needed to tread carefully. The 36-year-old son of the boss arriving with new technology to fix everything is enough to make any team roll its eyes.
So before building anything, I spent a few months doing the work myself – completing quotes, accepting orders, studying the data and asking questions.
The problem began to surface. Thirty years of rich data existed in the company’s database, and employees had decades of experience, but neither was helping the team work much faster or more strategically. Employees were buried in routine manual work, while the data was used mainly to record the past rather than inform the next decision.
When I asked my dad whether we could experiment with automation and AI in quoting, his reaction was appropriately sceptical. He did not care how impressive AI was. Anything we implemented had to reduce work and retain service quality without adding complexity.
Why quoting was the right place to start
When I first arrived, Delaware Ship Supply averaged about 104 minutes of human work to complete a 250-plus-line provisions quote.
Much of that involved scanning customer emails, downloading attachments, translating requests into internal inventory data, converting quantities and units, applying customer-specific pricing and returning the prices in the customer’s original format.
I started by automating data entry. Through a combination of modern workflow automation and targeted AI, I was able to turn external requests into consistently organised, usable data. I then used AI to translate messy customer item requests, quantities, units and special handling instructions into real inventory data, costs and prices. I built the system around a record of prior human decisions, with those decisions remaining the source of truth.
While AI handles inconsistent language, recognises patterns and resolves ambiguity, conventional software controls calculations, pricing rules, permissions and auditability.
Most importantly, the workflow separates routine processing from work requiring expert judgment. Instead of asking a quoter to recheck 250 lines, it flags the few where uncertainty, risk or commercial context calls for a person.
That experiment eventually became Breakwater.
Eventually, the same 250-line quote took about five minutes to complete. Roughly 97% of its lines were completed using accumulated memory and AI, with the remainder marked for human review. Not only was the data processing more automated, but the entire quote could be completed without ever opening Excel.
The team was not only quoting dramatically faster; it was also sending more quotes, responding to vessels on tighter timelines and winning more business.
The hidden economics
Over the next two years, we applied the same approach with ship suppliers of different sizes and operating models – from companies manually processing hundreds of quotes a day in modern ERP systems to teams already experimenting with AI themselves.
Although the technology varied, the underlying reality remained remarkably consistent: workflows that had changed little in decades, with people still doing too much of the routine work.
Across the implementations that have gone live to date, quote completion time has fallen by more than 90%.
From that work, three things became clear.
Quoting is expensive when humans do all of the work. Completing a quote means scanning portals and company inboxes, reading spreadsheets, entering data and translating customer requests into prices. By redesigning the workflow around the steps that truly require human judgment, a supplier can process more quotes and win more business in substantially less time.
Most suppliers already have more data than they can use. Years of data, costs, customer behaviour, prior quotes and commercial judgment are scattered across systems and people. But millions of data lines are not, by themselves, useful intelligence. A great AI development team can clean and organise that history. Only then can AI accurately put that data to work.
“Safe” technology has become riskier than ever. A well-known ERP may feel like the safest option. But a blue-chip logo can mask the cost of slow change, specialised hardware, ongoing IT support and expensive layers of add-ons that never fully fit the way a business operates. As AI makes sophisticated software more accessible, the technologies that earn their place will be purpose-built for ship supply and evolve quickly enough to turn new advancements in AI into practical value.
From faster quotes to an advantage that scales
When a quoter corrects something completed by AI – an item match, unit conversion, supplier choice or pricing exception – that judgment no longer disappears into a single transaction.
Each edit can improve future transactions by informing the system over time. In AI terms, this creates a continuous learning loop. Embedding it directly into daily work is what makes AI tools begin to feel like magic.
This turns an experienced employee’s judgment into institutional memory that future work can draw on. The same principle can extend into other workflows, including purchase-order acceptance, technical RFQ screening, procurement, cost updates from messy supplier PDFs and inventory planning.
A few suppliers have gone further and replaced ERP systems they once assumed they could never leave. That has allowed them to rebuild the entire quote-to-cash workflow around technology doing the routine work, while significantly reducing IT overhead and legacy-system costs.
The human impact has been just as meaningful.
One customer recently told me he was having more fun running his business than ever before. Another client’s most seasoned quoter with 35 years of industry experience told us that we helped him take his first ever two week vacation.
In the future, I believe ship suppliers will barely recognise today’s quote-to-cash workflow. Routine requests will move through pricing, purchasing, fulfilment and invoicing, with people stepping in for the decisions that require real experience. And there is a very real possibility of certain workflows happening fully autonomously.
The warehouse may be the next frontier. AI will help business leaders make near-real-time strategic decisions and predict demand, supply constraints and seasonality with far greater accuracy. I expect a more responsive and resilient industry in which people spend less time moving information and more time serving customers and vessels.
The companies best positioned for that future will not simply be those with access to the best AI model but rather those that are most willing to change from the inside out.
In the two years since I began working with my dad, Delaware Ship Supply has achieved financial results it had not previously seen in its more than 120-year history. AI was not the whole story, but it removed a major operational constraint and helped make it possible for my dad to retire this year.
The opportunity is not to replace what has made established ship suppliers successful. It is to make decades of their knowledge available in every transaction and build the foundation for what comes next.
Hearing a ship supplier who once told me, “I’m not too interested in AI,” now say, “It’s hard to believe this was possible,” says it all.
