Rising expectations expose procurement’s capacity crunch

Procurement leaders are being asked to deliver higher savings and broader business value while operating with limited capacity, leaving AI deployment, data improvement and supplier risk monitoring short of the maturity expected of the function.

ProcureAbility’s 2026 CPO Benchmark Study found that 61% expect to deliver more savings than last year, while 71% describe their team capacity as moderately or highly constrained. Some 42% face both conditions at once.

Being engaged too late was the leading barrier to savings, cited by 43%, well ahead of poorly managed contracts at 19%. Yet only 26% of CPOs identified late involvement as a problem, compared with 55% of directors and category managers.

Cost savings remain the dominant measure of procurement performance, used by 60% of respondents. Risk mitigation followed at 43% and working capital improvement at 39%, while supplier-driven innovation registered 22%, revenue impact 15% and speed to market 14%.

AI adoption has spread, but scale remains rare. Some 40% are running pilots and 44% have limited deployments, while just 11% have scaled AI across multiple processes. Reported value rises with maturity: 67% of scaled users see moderate or significant value, compared with 50% at limited deployment and 35% among pilot users.

“It may be tempting to skip steps in the rush to adopt AI, but doing so creates real risk,” said Satyen Pathak, managing director for digital and AI at ProcureAbility. “Organizations that move too quickly without the right foundations in place are likely to face costly setbacks later.”

Data quality was cited by 50% as a barrier to scaling AI, followed by legacy technology at 49%, skills at 48% and governance at 39%. Uncertainty over returns ranked last at 26%, suggesting readiness is a greater obstacle than confidence in AI’s potential.

Usage also remains concentrated in task-based applications. AI copilots are used by 61%, contract analytics by 60% and spend analytics by 59%. Adoption drops to 41% for sourcing automation and 32% for supplier risk monitoring.

A wider data gap compounds the problem. Although 70% use analytics frequently or constantly, only 38% rate their capability as advanced and 51% name data quality as their biggest analytics weakness.

Risk monitoring shows an even sharper mismatch. Geopolitical exposure concerns 62%, logistics disruption 55% and regulatory change 51%, yet only 6% have predictive or real-time monitoring. Another 62% still rely on manual or periodic tools.

Supplier relationships offer little protection. Some 68% expect supplier-driven innovation, but just 14% report genuine strategic partnerships and 53% remain at transactional or basic performance management. Supplier consolidation is used by 46% as a savings lever even as 47% identify concentration as a risk.

ProcureAbility recommends earlier involvement in planning and budgeting, stronger data governance, organisational preparation for AI, workforce upskilling and closer collaboration with critical suppliers. That agenda reflects a function whose ambitions have expanded faster than its ability to execute them.

The survey covered 160 leaders across 21 industries: 64% in North America, 8% in Europe and 28% elsewhere. Sixty per cent held CPO, vice-president or director roles, with annual spend from $50m to over $10bn.

 

US inflation revision shifts Fed hike outlook to December

Goldman Sachs Research has pushed its forecast for the Federal Reserve’s next interest-rate increase from October to December after methodological revisions lowered the recent path of core US inflation more sharply than expected.

Core personal consumption expenditures inflation rose 0.25% month on month in August, below Goldman’s 0.27% forecast and the 0.3% consensus estimate. Its annual rate reached 3.01%, against expectations of 3.3%.

July’s annual core reading was revised down by 0.36 percentage points to 2.98%. Changes to the treatment of portfolio management accounted for 0.23 points of that reduction, while computer accessories and software removed another 0.11 points. A small upward revision to legal services offset part of the decline.

Those changes led Goldman to cut its 2026 fourth-quarter forecast for core PCE inflation to 3%. That would undershoot the median Federal Open Market Committee participant’s projection of 3.4%.

“Coupled with comments from New York Fed President John Williams yesterday, we now think that an October hike is unlikely,” Goldman Sachs Research said. “We are pushing back the second hike in our forecast to December, and we see a strong chance that the FOMC will ultimately conclude that additional rate hikes are unnecessary.”

Headline PCE inflation offered a less benign reading, rising 0.31% during August and 3.42% from a year earlier. Both measures were close to Goldman’s estimates, although the annual increase was below the 3.7% consensus forecast.

Economic activity remained firm enough to keep a December move in consideration. Real GDP growth for the second quarter was revised up by 0.7 percentage points to an annualised 2.2%, reflecting stronger consumption and investment. Consumer spending growth was lifted to 3.8%, while investment growth was revised to 4.6%.

Annual revisions also raised first-quarter GDP growth by 0.4 points to 2.5%. Changes to earlier years were modest, leaving 2025 growth unchanged at 2% while lowering 2024 by 0.2 points to 2.4%.

August personal spending increased 0.9%, while personal income grew by a weaker-than-expected 0.2%. Disposable income was 4.8% higher than a year earlier, and the savings rate fell by 0.5 points to 4.1% after substantial revisions to previous estimates.

Trade supplied the weaker element of the growth picture. The goods deficit widened by $13.7bn to $132.6bn, compared with the $115.3bn consensus forecast, as a $17.4bn increase in imports outpaced a $3.7bn rise in exports.

Private employment provided a further sign of resilience, increasing by 90,000 in September against expectations of 75,000. Services added 59,000 jobs and goods-producing industries 31,000.

Goldman nevertheless trimmed its third-quarter GDP tracking estimate by 0.1 points to an annualised 3.3%, principally because stronger imports outweighed firmer spending and inventory data. Softer measured inflation has weakened the case for an October hike, but resilient demand leaves the December decision open.

 

BHP trials tokenised deposits across borders

Global mining giant BHP has tested tokenised deposits in live US dollar payments between Melbourne and New York, exploring whether global treasury funds could eventually move across borders in real time without changing the banking channels used by corporate teams.

The transaction involved ANZ, BHP, Citi and Swift. ANZ described it as one of the first live corporate treasury payments conducted with tokenised deposits through Swift’s blockchain-based ledger and the first use of that infrastructure by an Australian bank.

BHP initiated the payments from its existing ANZ account and channel. Tokenisation and distributed ledger technology operated behind the scenes, allowing the company to follow its established banking process rather than adopt a separate interface or workflow.

Swift’s shared ledger connected the participating banks and enabled tokenised deposits issued by different institutions to interact. The test examined how digital settlement infrastructure could work alongside existing bank accounts, controls and operational arrangements.

Its central attraction for corporate treasury is the potential to process cross-border payments faster across time zones and outside conventional banking hours. Shortening the wait between releasing funds and confirming their arrival could give companies greater flexibility over when liquidity is transferred between markets.

Marlon Singh, manager for treasury at BHP, said: “The opportunity we see is the ability to move funds across borders in real time. Rather than waiting for banking systems to open, global businesses could eventually transact around the clock with greater certainty and efficiency. This pilot with ANZ and Swift gives us the opportunity to test whether emerging payments technology can help make that a reality. While still at an early stage, initiatives like this have the potential to simplify liquidity management and make it easier for us to move capital when and where it is needed.”

Interoperability will be central to any wider use. Corporate payments regularly cross several institutions and infrastructures, making the ability of separately issued tokenised deposits to operate through a common ledger more important than a payment confined to one bank’s network.

Keeping the underlying process invisible could also reduce the operational change required from corporate users. Existing accounts and channels would remain the point of access, with tokenisation handling movement between participating banks.

The transaction provides a live test of how tokenised commercial bank money could fit into established treasury processes. Its broader significance will depend on whether the approach can be repeated across more banks, currencies and operating hours while retaining the controls and certainty companies already expect.

 

Kyriba connects treasury data with enterprise analytics

Kyriba has launched a data service that allows companies to move structured treasury information into the business intelligence and data platforms used by finance and IT teams. Data-as-a-Service, or DaaS, provides governed access to treasury data for high-volume analysis, historical reporting and integration with wider enterprise information. It is intended to reduce file downloads, manual reformatting and separate data processes created for individual reports.

The service connects with existing BI, reporting and data platforms through JDBC, OAuth2 and SQL. Incremental refreshes allow companies to update datasets without repeatedly extracting their entire history.

Customers can access more than 165 tables and 4,400 fields spanning cash, payments, financial risk and supply chain finance. Up to three years of historical information will be available from the outset, subject to the data held for each customer.

That depth could allow treasury information to be analysed alongside sales forecasts, operating expenditure, purchase commitments and workforce plans. Finance teams could then examine how changes elsewhere in the business may affect liquidity, working capital or forecast cash positions.

Monica Boydston, chief product officer at Kyriba, said: “This gives businesses a more connected view of financial performance and creates new opportunities for enterprise decision-making.”

Treasury data has traditionally been gathered and managed for activities such as daily cash positioning, payment execution and risk oversight. Making it available within enterprise analytics environments could extend its use into financial planning, performance analysis and scenario modelling.

A sales forecast could, for example, be assessed alongside expected receipts and currency exposures, while procurement commitments could be viewed against available liquidity and scheduled payments. Historical information may also help companies compare forecast assumptions with previous cash movements.

IT teams retain a role in controlling how information enters company data environments. Using established connectivity and governed datasets may reduce the number of bespoke feeds they must maintain when departments request access to treasury records.

The launch reflects a broader shift towards treating treasury records as shared financial data available beyond a specialist system. Its practical value will depend on how effectively companies combine those records with operational data and turn the analysis into financing, liquidity and working capital decisions.

 

BMO and Mastercard embed virtual cards in ERPs

BMO and Mastercard have introduced embedded virtual card payments that can be initiated, approved and reconciled from corporate software platforms in Canada and the US. Eligible BMO Corporate Card clients can access the capability through participating ERP, procurement, accounts payable and travel systems. Companies can make supplier, travel and expense payments within the software used to manage them.

The service uses Mastercard’s Commercial Express framework to connect platform workflows with BMO-issued virtual cards. BMO said it is the first Mastercard issuer in Canada to offer this integrated experience through participating software providers.

Embedding execution into enterprise systems can remove the hand-off between finance software and a separate banking portal. A payment can move from approval to card issuance and reconciliation within the same workflow, reducing manual intervention and transfers of data between systems.

Virtual cards can carry controls governing how, where and for how much they are used. Transaction information generated alongside the payment can support automated reconciliation, giving accounts payable and treasury teams a clearer link between an approved obligation and its settlement.

Diane Miquelon, senior vice-president for financial institutions partnerships at Mastercard Canada, said: “As businesses digitise their operations, payments must move seamlessly alongside day-to-day business activities.”

For procurement and accounts payable teams, the integration places payment execution closer to invoice approval. Travel and expense platforms can similarly generate controlled payment credentials within their existing processes.

Treasury retains visibility over card-funded payments while operational teams complete approved transactions from their usual systems. The model may also help companies time payments more precisely and retain the working capital characteristics associated with commercial cards.

Participating software providers gain a common framework for connecting to issuer-backed virtual cards, reducing the need to build a separate integration for each payment arrangement. That standardisation could support wider availability as additional platforms connect to Commercial Express.

Access is currently limited to eligible BMO clients and software platforms participating in the integration. The launch nevertheless reflects the continuing movement of commercial payments away from stand-alone bank channels and into the ERP, procurement and accounts payable environments where companies create and approve their obligations.

 

HSBC connects corporate AI tools to banking data

HSBC has launched a digital banking service that allows corporate and institutional clients to connect their own AI tools to authorised account and transaction data. HSBCnio combines transaction banking functions with web, mobile and direct system connectivity. Clients can use the interface to view cash balances and transactions, track payments, arrange trade loans and access foreign exchange services.

Its most distinctive element is access through HSBC’s Model Context Protocol, which enables customers’ AI tools to retrieve permitted banking information. Existing permissions and controls will govern what data those tools can use.

The approach could allow a company to bring bank data into AI-supported treasury workflows without first moving it through spreadsheets or creating a separate information source. Possible applications include investigating account movements, following payment status and bringing current banking information into cash or liquidity analysis.

Amber Henderson-Smart, global head of client connectivity at HSBC Global Payments Solutions, said: “The new solution brings transaction banking capabilities into the channels and workflows that clients use today, while helping them integrate AI through secure, permissioned access to their banking information.”

Customers will be able to choose between working through HSBCnio’s web or mobile channels, connecting corporate systems directly or using authorised AI tools. That range reflects the different ways treasury teams access banking services, from individual users reviewing transactions to systems consuming data automatically.

HSBC also plans to add an AI-enabled feature called Ask HSBC. It is intended to let users query and interpret their banking information, although the bank described this as a future addition rather than part of the initial capability set.

The launch brings AI connectivity closer to day-to-day banking activity. HSBCnio creates a permissioned route to source information within the banking environment, extending AI use beyond analysis performed after data has been exported.

Practical adoption will depend on how companies govern access by their AI tools and fit the resulting interactions into existing approval, security and audit arrangements. For finance teams already experimenting with AI, the immediate development is access to live authorised banking information through the protocol alongside conventional web, mobile and system connections.

 

Citi and Coinbase link stablecoin and fiat payments

Citi and Coinbase are connecting bank-based payments with stablecoin infrastructure through two services aimed at corporate payment collection, account management and merchant checkout. Coinbase Virtual Accounts will use Citi’s Virtual Account Wallet. Delivered through Citi Services’ Banking-as-a-Service infrastructure, the arrangement gives Coinbase payments customers account-like functionality for accepting, holding and paying funds.

Incoming fiat payments will be converted automatically into stablecoins. This removes the need for customers to operate separate banking and digital-asset processes when moving funds onto blockchain-based payment rails.

A second initiative will allow institutional clients using Spring by Citi to accept stablecoins during online checkout. Coinbase Payments will process the digital currency, which will be converted into fiat automatically before Citi settles the funds as bank of record.

Merchants will therefore be able to offer stablecoin payments without holding, safeguarding or managing the digital assets themselves. Citi and Coinbase said the capability could give participating businesses access to payments from more than 150 million stablecoin holders worldwide.

Both services will launch initially in the US. Further capabilities are planned as Citi and Coinbase extend their work across conventional and blockchain-based payment systems.

These payment flows address different sides of the conversion between fiat money and stablecoins. Coinbase Virtual Accounts will turn incoming conventional currency into digital assets, while the merchant service will convert customers’ stablecoins into fiat for settlement.

For corporate users, that distinction matters. Businesses receiving payments through Coinbase may want funds available on digital rails, whereas merchants accepting stablecoins may prefer their proceeds to arrive as conventional bank money without adding digital-asset custody to their operations.

Citi’s involvement also keeps the banking layer within the transaction. Its virtual account infrastructure will support the Coinbase service, while its role as bank of record places fiat settlement for merchant payments within an established bank relationship.

Embedding the exchange between fiat and stablecoins into existing payment products is intended to reduce the operational divide between bank accounts and blockchain networks. Adoption will depend on customer demand for stablecoin payments and how readily the services fit existing treasury controls, reconciliation and liquidity processes.

 

BofA expands real-time payments across treasury and borders

Bank of America is extending real-time payments across two distinct corporate needs: high-value domestic liquidity movements and lower-value international transfers delivered through local instant-payment networks.

Corporate use of the US RTP network provides the clearest sign of adoption. Bank of America clients increased transaction volumes by 48% between January and July compared with the same period in 2025. Transactions worth more than $1m rose 351%.

Growth followed an increase in the network’s individual payment limit from $1m to $10m in February 2025. That change expanded its use beyond smaller urgent payments into intercompany transfers, cash concentration, supplier payments and movements between corporate accounts at different banks.

Round-the-clock availability allows companies to reposition funds after normal banking hours, fund subsidiaries before the next business day and adjust cash positions during month- or quarter-end close. Internal sweeps can also be completed when liquidity is required rather than held until the next processing window.

A pilot with Brazilian bank Bradesco applies real-time processing to a different part of the payment market. Bradesco initiated a cross-border transaction through its existing Swift connection, with the funds delivered in Hong Kong dollars to a local beneficiary through Hong Kong’s Faster Payment System.

Bank of America’s cross-border solution is designed for high-volume, lower-value transactions. It provides payment tracking, certainty over delivery and full-principal receipt, while using established bank connectivity for initiation and a domestic instant-payment network for the final leg.

Future access through CashPro will give corporate and institutional clients another route for initiating these payments. Bradesco and Bank of America are also testing a separate US payment route through Swift for consumer and small-business international transfers.

That combination illustrates how real-time payment adoption is dividing into use cases with different values and purposes. Higher domestic limits support liquidity management and intercompany funding, while cross-border connections target frequent payments that benefit from faster local delivery and improved tracking.

Across both models, companies gain greater control over timing. Payments can be released closer to when funds are required, reducing dependence on conventional processing windows and giving finance teams more options for managing liquidity across accounts, entities and markets.

 

TIS weighs buy versus build for AI cash forecasting

Treasury Intelligence Solutions has published a white paper comparing the cost and operational demands of building AI-powered cash forecasting internally with buying a specialist platform. 'Buy vs. Build: The Real Cost of Building AI-Powered Cash Forecasting In-House' divides organisations into three complexity profiles based on their entities, banking and ERP connections, revenue and engineering requirements.

At the lower end, TIS models a company with one to three entities, two or three integrations and one full-time engineer. It estimates an internal build would take nine to 12 months and describes this route as potentially viable, although carrying execution risk.

Its medium-complexity profile assumes five to 10 entities, eight integrations and two engineers, with development taking 12 to 18 months. A high-complexity organisation has at least 15 entities, more than 18 integrations and three or more engineers, pushing the estimated timetable to 18 to 24 months.

Comparing those schedules, TIS gives its own platform a four-month deployment timetable. Its three-year framework considers initial development, bank and ERP connectors, data cleansing, multi-entity consolidation, maintenance, model retraining, feature development and reliance on scarce technical staff. It assigns qualitative cost ratings rather than monetary estimates.

Charles Bennett, chief product officer at TIS, said: “The question isn’t whether organisations can build AI-powered cash forecasting capabilities. Many can. The more important question is whether maintaining that capability is the best use of treasury, finance and IT resources over the long term.”

The paper closes with questions designed to test whether an internal build remains economical beyond launch. It highlights the developer time needed to maintain bank and ERP connections, the cost of a six-month delay and the risk that crucial knowledge leaves with a handful of employees. Expansion can add further strain, as each additional entity, bank or country brings more data, integrations and maintenance. The decision therefore rests as much on the resources needed to keep a forecasting system current as on the cost of building its first version.

 

Amex links corporate cards with expense management

American Express has combined a corporate cashback card, expense-management software and a mobile app in an updated offering for business customers.

The Corporate Cashback Card provides 1.5% cashback on eligible purchases up to $10m per company each calendar year, falling to 1% above that threshold. Its annual programme fee is $2,950, waived for the first year, and covers unlimited physical and virtual cards.

Finance teams can set controls governing where, when and how individual cards are used. Transaction data flows into Amex’s expense software, which also supports tailored submission requirements, approval rules, statement access and integrations with widely used ERP systems.

Employees using the Amex Expense app will have draft expense submissions generated when purchases are made. AI recommends fields based on previous expense history, checks uploaded receipts against transaction details and compares claims with company policy before approving them or sending them for review.

American Express plans to add two AI agents later this autumn. One will contact employees through SMS or Slack to collect missing information, suggest classifications and submit completed expenses. A second will allow finance teams to question company spending data and generate supporting charts.

Accounts payable software is also planned, covering invoice intake, approvals, payments and reconciliation. American Express said this would allow customers to manage supplier bills alongside employee expenditure, although it did not give a launch date.

Availability is initially limited to new Corporate Cashback Card customers. Spending capacity will be determined by each company’s finances and payment history, rather than a fixed limit.