# Spend Management's Agentic Turn: AI Cost Controls and MCP — July 2026 > Ramp, Mercury, Spendesk and Payhawk all shipped agent-facing infrastructure this period, moving spend management competition from expense capture toward autonomous financial operations. *Spend Management & Finance Automation · general · 18 July 2026* Spend management vendors spent this period wiring AI agents into the ledger rather than around it. Ramp extended spend controls to AI token consumption, Mercury and Spendesk shipped Model Context Protocol layers for querying live financial data from third-party assistants, and Payhawk pushed native SAP S/4HANA integration into the enterprise. The competitive line is shifting from card issuance to which platform an autonomous agent can safely operate. ## Key Findings - Ramp extends spend management to AI infrastructure: Token Spend Management and the Ramp Router open its internal multi-model LLM routing to customers, targeting what Ramp identifies as the fastest-growing category of business spend. - Mercury launches CLI for AI agents: This move targets developers and AI-based startups, moving banking beyond traditional UI-based interactions. - Spendesk achieves European profitability: Reaching this milestone ahead of schedule allows the firm to pivot from venture-fueled growth to sustainable mid-market competition. - Navan acquires Smartrips: The purchase of the Sao Paulo-based TMC secures direct inventory in Brazil, which accounts for roughly 40% of Latin American travel spend. - Expensify pauses recruitment: Zero listed vacancies following a Q1 revenue shortfall against forecast suggests a significant strategic restructuring. - Payhawk integrates native SAP S/4HANA: The Summer '26 update targets enterprise complexity by removing third-party middleware for financial reconciliation. - Bill defends on service friction: Verified reviews log $90 payment-reversal fees and automated closure of six-figure credit lines, even as Bill repositions its brand around agentic AI. ## Ramp vs Bill in 2026 Ramp is pressing its advantage over Bill by extending spend management into territory Bill has not claimed: the cost of AI itself. Ramp's Token Spend Management tool tracks AI consumption across major providers, and its Ramp Router exposes internal multi-model LLM routing to customers as an OpenAI-compatible endpoint. That positions Ramp as infrastructure rather than a card programme. Bill, meanwhile, is defending against documented service friction, verified reviews log a $90 fee for a single payment reversal and the automated closure of $100,000 credit lines, alongside recurring reports of support latency. You can analyze these shifting dynamics further in our [Bill vs Ramp](/compare/bill-vs-ramp) comparison guide. While Bill maintains a deep proprietary payments network, it is also contending with a technical anomaly in which users reported high-volume decline attempts on cancelled or replaced corporate cards. Bill's answer has been a visible repositioning around AI-driven automation, including an autonomous W-9 agent that handles vendor outreach, collection and validation without a human in the loop, and a multi-entity inbox for consolidated document intake. The contest between the two is increasingly about which platform an autonomous agent can be trusted to operate. ## Mercury AI agent banking launch July 2026 Mercury is pivoting its product strategy toward the 'agentic economy' by introducing a Command-Line Interface (CLI) specifically designed for AI agent workflows. This allows businesses to interact with their bank accounts programmatically, moving beyond the limitations of traditional APIs. This technical shift is supported by a $200 million Series D funding round that values Mercury at $5.2 billion, following four consecutive years of profitability. The firm is leveraging this capital to develop 'Mercury Command,' a natural-language interface for financial operations, serving a base of 300,000 businesses. Despite this innovation, Mercury faces growing user dissatisfaction regarding compliance handling. Its rating has slipped to roughly 3.3 stars against a historical average near 3.84, with complaints clustering on automated document rejection and account closures. This creates a reliability gap that competitors like [Ramp and Brex](/compare/ramp-vs-brex) are positioned to fill: Brex, for instance, points to a documented 7-day reduction in month-end close for the Boston Celtics. ## Why companies are switching from Expensify to Spendesk Spendesk is successfully displacing Expensify and Pleo by emphasizing multi-entity budget control and reaching a critical profitability milestone ahead of schedule. User sentiment indicates that Spendesk’s 'Budgets 2.0' update, which uses AI to automatically build budget structures from spreadsheets, addresses the manual tracking pain points that plague Expensify users. Expensify is currently in a defensive posture, reporting a Q1 revenue shortfall of $34 million and a net loss of -$0.02 EPS. Furthermore, Expensify users have reported critical failures in QuickBooks Desktop sync following Windows 11 updates, leading to 1-star ratings for the interface. For a deeper look at European alternatives, see our [Pleo vs Spendesk](/compare/pleo-vs-spendesk) analysis. While Spendesk gains ground, it is not without its own integration vulnerabilities; mid-market users have reported that NetSuite integrations can be a 'manual mess,' and some SMBs have cited 10-day delays in accessing funds during account off-boarding. However, Spendesk’s launch of a Model Context Protocol (MCP) for real-time querying via Claude and ChatGPT has positioned it as a more forward-looking infrastructure layer compared to Expensify’s current hiring freeze and zero active job listings. ## How Navan and SAP Concur are evolving in 2026 Navan is expanding its global footprint through the acquisition of Smartrips in Sao Paulo, targeting Brazil, which accounts for roughly 40% of Latin American travel spend, its first purchase since the October 2025 IPO. Navan has also moved into the MICE market with Navan Events, integrating venue sourcing with attendee travel and financial reporting. However, Navan faces criticism over price volatility, with users reporting flight price increases of $150 to $500 while waiting for manual manager approvals within the app. SAP Concur is defending its enterprise dominance by embedding its Joule AI copilot and autonomous Receipt Analysis Agents into T&E workflows, part of an SAP-wide goal of automating a large share of core business tasks. The modernisation targets a well-documented weakness: reviewers consistently describe the interface as 'dated' and 'click-heavy', and manual hotel itemization remains a friction point among a reviewer base that is roughly 70% enterprise. Meanwhile, Payhawk has shipped its 'Agent Fetch' autonomous retrieval tool globally, claiming a 46% reduction in invoice submission times and retrieval roughly 12x faster than manual processes. Qonto and Tipalti continue to focus on compliance and 'agentic commerce'. ## Frequently Asked Questions ### What are the latest product updates from Ramp in July 2026? Ramp launched 'Stack,' an AI-powered accounting operating system designed for top-tier CPA firms, alongside a Q2 product suite featuring AI purchasing agents and token controls. Additionally, Ramp released a Model Context Protocol (MCP) for direct connectivity with Perplexity AI and integrated with Accounting Seed for Salesforce-native reconciliation. ### Why are companies switching from Pleo and Expensify to Spendesk? Recent user sentiment analysis shows Spendesk is displacing Pleo and Expensify by leveraging its new AI-automated budgeting structure and real-time querying capabilities via the Model Context Protocol (MCP). Spendesk's focus on AI-driven financial workflows has positioned it as a more modern alternative to legacy spend tools. ### How do Ramp and Mercury compare in terms of valuation and AI strategy in 2026? Ramp reached a $44 billion valuation following a $750 million Series F, focusing its AI strategy on 'Stack' for CPAs and Perplexity AI integrations. Mercury is currently valued at $5.2 billion after a $200 million Series D and is targeting developers with a new Command-Line Interface (CLI) specifically for AI agent workflows. ### What is the Model Context Protocol (MCP) and which B2B spend platforms are using it? The Model Context Protocol (MCP) is a new standard for real-time data querying by AI assistants, adopted this period by Ramp, Spendesk, Qonto, and Navan. These integrations allow third-party AI agents to interact directly with financial data for autonomous workflows and conversational analytics. ### How is AI disrupting B2B spend management according to recent 2026 reports? Disruption is driven by autonomous agents, such as Payhawk’s 'Agent Fetch' for invoice collection and Pleo’s autonomous financial workflows. Companies like Mesh Payments are introducing AI orchestration layers for ERPs, while Brex has demonstrated a 7-day reduction in month-end close for enterprise clients like the Boston Celtics using automated vendor sync. ### Is Ramp targeting Bill.com customers with specific displacement campaigns? Yes, Ramp has initiated targeted marketing campaigns specifically designed to displace Bill (formerly Bill.com) by highlighting superior workflow efficiency. These campaigns emphasize Ramp's new AI purchasing agents and direct ERP connections as key advantages over Bill's current platform. --- Source: IndustryLens — automated competitive intelligence. Read online: https://industry-lens.com/reports/spend-management-s-agentic-turn-ai-cost-controls-and-mcp-july-2026 Competitors monitored: Bill, Brex, Expensify, Mercury, Mesh Payments, Navan, Payhawk, Pleo, Qonto, Ramp, SAP Concur, Spendesk, Tipalti.