Trade finance and B2B payments are on a path of increased automation of customer onboarding and due diligence for financing approvals, disbursements and repayments. The new wave of generative artificial intelligence such as ChatGPT is supercharging this trend. In the process, it’s freeing up and lowering the cost of working capital.

Trade finance is to global trade what water is to life. The questions are: Who finances the trades? With what capital? How much is the cost of financing? And who bears the risk?

With the unmet demand for financing exceeding $2 trillion in 2022, the trade finance gap is the “elephant in the room.” A crucial impediment to sustainable and equitable growth today, the problem affects small and medium-sized suppliers and buyers disproportionally.

The catalogue of trade finance products is long, cluttered and heavy on jargon. Invoice factoring, purchase order finance, dynamic discounting, supply chain finance, flexible payment schemes: all have their merits and tailored benefits to counterparties. But such offerings also add complexity, obfuscate the risks involved, and increase processing costs.

Decluttering the trade finance space is a challenge. Yet it’s essential in order for generative AI to operate effectively.

Consider trade finance as the union of two “chains”: the physical and the financial. Physical supply chains move goods downstream to end customers. Financial chains move cash upstream to suppliers. Simply put, financing can occur at any point of the procurement cycle — from establishing a purchase order, at product release, when shipping overseas, or even at customer invoicing with net payment terms.

Irrespective of the timing for financing, the counterparties always need to establish the financing amount, security or collateral value, and cost for financing. The core of any trade financing offering is the same; the jargon only reflects the “packaging.”

The main concept underlying the two chains is risk, which is best understood in two dimensions —credit and performance. Financiers are typically more comfortable with credit risk, as in “What are the chances that the buyer won’t be able to pay on time and in full, due to a deteriorating balance sheet?” Suppliers’ strong suit lies in controlling the performance risk, as in “What are the chances that the supplier won’t be able to fulfil a purchase order on-time and in full?”

Note that as procurement milestones are met from order to delivery and invoicing, performance risk decreases and gives way to credit risk. 

Mind the $2 Trillion Gap

So why the $2 trillion mismatch between supply and demand for trade finance?

When it comes to challenges, trade finance is a mirror image of global supply chains. It’s document-heavy and paper-based, manual and labor-intensive, consisting of multiple data silos, total lack of transparency and many disconnected systems. Add to that the complexity of historically disparate functions — supply chain and finance — needing to operate in unison, yet with distinct mindset and objectives.

Such inefficiencies are a result of the current toolkit and software, which make the cost of financing any application expensive. Because applications for small and medium-sized businesses can’t be processed profitably, larger financial institutions won’t work with them. Good software is lacking, so the need for trade finance digitization and automation is urgent. 

Consider an importer of pharmaceuticals in emerging markets with manufacturing overseas, a high-value drug product, and a tightly regulated supply chain. To fulfill purchase orders for local distributors, access to fast and fair working capital is essential — four to nine months ahead of delivery or final payment.

The typical end-to-end trade finance process includes a maze of tedious processing steps. They include customer onboarding, “know your customer” (KYC) and due diligence, anti-money laundering (AML) clearance, commercial trade documents, insurance and regulatory documents, shipping documents, risk underwriting, contracting and lien filing, cross-border payments, trade invoicing, trade monitoring, repayments and reconciliation.

Self-Driving Trade Finance

Next-generation automation, delivered through a software-as-a-service (SaaS) model, can dramatically drive down the cost to process an application, freeing up cash where it’s needed most: at the source, for suppliers and exporters to secure products and services in a timely manner. Imagine B2B software that can approve financing, disburse cash and reconcile repayments, all autonomously via generative AI. Enterprise AI acts as a trade finance expert — in superhuman form.

A report by Deloitte notes that venture capital investments in generative AI exceeded $2 billion 2022, while Gartner estimates that more than 10% of all data will be AI-generated by as early as 2025.

Enterprise AI consists of three defining elements:

• Large language models. LLMs such as ChatGPT understand language and generate text when given a prompt. Model parameters number in the billions or even trillions. They are typically trained in massive amounts of unlabeled data, such as that derived from the internet, as well as proprietary enterprise data.
• AI agents. The software senses its environment, autonomously makes decisions, takes action, senses the outcome and iterates to optimize performance toward the objective.
• Sequence chains. These consists of a sequence of steps that an AI agent autonomously performs to accomplish a more complex objective. They may include prompting an LLM, fetching data via an application programming interface (API), or utilizing any available digital tool, whether public or proprietary.

Value from these tools is realized in three steps: digitization, automation, and de-risking.

The foundation is the digitization of the end-to-end trade finance journey, involving electronic documents such as POs, financial statements, KYC documents, bills of lading, invoices and e-mails, as well as physical and financial transactions such as shipping milestones, contract execution, and wires or automated clearinghouse (ACH). 

AI agents are trained to sort, label, parse and interpret commercial, legal, shipping, and other documents and data to stitch together a unified, digital view of the supply chain, commercial transactions, and financing options. They ask: Is the PO profit margin financeable? Is the net interest margin (NIM) sufficient? Are documents current, correct and valid?

After that comes workflow automation, with streamlined onboarding, API-driven AML clearance, cash advances and direct debit repayments, and automated tracking of orders, shipments and payments. 

AI agent-assisted KYC and due-diligence workflows use the company name to aggregate, interpret and cross-reference information from self-reported, proprietary, public, and third-party data such as open corporates, open banking, AML sanctions and newsfeeds. They then synthesize summarized, structured business intelligence output, with positive and negative sentiment narratives and scoring tailored to trade finance and risk for prospective borrowers.

De-risking already starts with built-in process compliance through digitization and automation. It extends to credit and performance risk evaluation based on trade history and trends, as well as real-time identification of exceptions and proactive alerts.

Time to Take Off the Seat Belt?

Assembling a holistic experience with a high degree of automation is not a trivial matter. It involves integration across disparate systems, such as order and transportation management, payments and accounts receivable, through the use of well-trained generative algorithms.

Enterprise-specific generative AI models are still in the exploratory phase — the need for a human in the loop isn’t going away anytime soon. For edge cases where the algorithm has low confidence or fundamentally fails, the outcomes still need to be manually reviewed and validated.

Data is the true enabler. Pre-trained models and publicly available transformers are novel concepts, and of significant value if put to good work. However, they can — and soon will — be commoditized. The true differentiator is the use of enterprise datasets for case-specific fine tuning and training. 

Trade finance and B2B payment automation by next-generation cloud-based software, enriched with enterprise AI, can lead to better scalability, profitability and stability. What’s more, some of the supply chain’s “unsung heroes” — such as forwarders, customs brokers, and small exporters and importers — have a lot to gain.

Are Freight Forwarders Bankers?

They are, but they don’t want to be. Most forwarders are unwillingly financing their customers and eating the cost because their customers — the shippers — have the negotiating power. Typically, a shipper wants to place an order for freight services, secure container capacity, have its shipment delivered, then pay the freight invoice at delivery or even 30-60 days later. Carriers, meanwhile, demand freight payments up front, before customs clearance or delivery.

Forwarders are forced to hold this “hot potato’ by paying for freight up front, and extending payment terms to their customers. Even if it’s for a short 30-90 days, the cost of financing piles up quickly, locking up cash and straining balance sheets.

What if forwarders could turn this cost center into a profit center, by giving net terms and financing options to their customers for freight services? Or even utilize excess cash flow to finance goods for the customers who opt in?

The answer is embedded trade finance — payments and finance workflow as a service, built into shipment-booking tools or other transportation management systems. Thousands of freight transactions each day can only be processed automatically. Manual intervention is only needed to manage exceptions. 

How to Get Started

The trade finance status quo is onerous and cost-prohibitive, leaving SMBs empty-handed, and preventing them from market entry, equitable growth and sustainable development.

So what’s a pragmatic approach for finance and supply chain leaders to get started?

• Adopting the right business model. Understand your role in the trade finance marketplace, and define your strategy accordingly. Are you a forwarder just seeking to offer net terms for freight payments to your shippers? Or would you like to provide options for goods financing as well? Are you willing to put your own working capital to work, or are you seeking external capital?
• Segmenting your customers. Understand your business segments and individual customers in terms of volume, profitability, and especially risk — both credit and performance. Segmentation creates a solid foundation to build upon, as entities can be pre-approved and underwritten efficiently for high-speed financing.
• Starting small. Select an enabling technology partner and identify a handful of your customers in need of financing. Pilot an embedded finance tool with limited automation to retain processing controls, then progressively “let go of the wheel” as you progress toward the self-driving mode.

Nikos Papageorgiou is co-founder and chief operating officer of Fishtail.ai.