This week Gartner put a number on something many managers already suspected: USD 234 billion in enterprise software spending is at risk due to agent AI. And in parallel, SAP announced that its AI Agent Hub is arriving in Q3. The two news items tell the same story: the AI agent government it stopped being an architecture topic to become a management problem. This week we reviewed 21 signs on five fronts, and they all point to the same gap: we adopt faster than we govern.
1. SAP News: AI agent governance becomes a product
[Alto] SAP Business AI Platform: BTP, Business Data Cloud, and BTM Consolidate
In Sapphire 2026, SAP announced the consolidation of BTP, Business Data Cloud, and Business Transformation Management under a unique three-layer architecture: context, build with Joule Studio 2.0, and govern with AI Agent Hub. The move acknowledges a real problem: fragmentation was generating more pilots than results.
For the business, this means that agents automatically inherit business context, data lineage, and process rules without manual configuration. The partner competition «SAP BTP» was renamed «SAP Business AI Platform» as of June 30. SAP is reorganizing its entire ecosystem around AI. View analysis.
[Alto] SAP AI Agent Hub available in Q3 2026
Built on SAP LeanIX, the AI Agent Hub will be GA this quarter. It functions as a centralized control plane to discover, verify, observe, and optimize all agents in the environment. It allows for defining architectural boundaries, data access policies, and monitoring against KPIs.
It is SAP's answer to a question most companies aren't asking yet: who audits the agents that are already making decisions. For an IT manager, this component is more strategic than any new ERP functionality. The AI agent government it is becoming a product category, not an optional best practice. View details.
[Alto] SAP Business One: roadmap 2026-2027 with Fiori Web Client and Document AI
SAP published the Business One roadmap for 2026 and 2027. The current version remains 10.0 (SQL and HANA), with the Web Client rebuilt on SAP Fiori principles: cleaner navigation and consistency across modules.
Three capabilities with direct operational impact stand out: SAP Document AI for automatic vendor invoice capture, generative AI-powered query generation with «Ask GenAI,» and Process Maps for visualizing flows and detecting bottlenecks in real time. Version 11 will arrive in 2027 with a deliberately short overlap. Those who do not plan their migration now will suffer the consequences later. View roadmap. If you want to dive deeper into how to get real value from the platform, check out Maximize SAP Investment.
[Medium] SAP Business One 10.0 FP 2605: Native Federated Authentication and 2FA
Feature Pack 2605 incorporates support for SAP's identity and authentication management service, allowing authentication against the corporate identity provider and enabling two-factor authentication. For an industrial SME, this eliminates one of the most common security holes: ERP users with local passwords without a corporate policy.
The impact is not just technical. It is about compliance and operational continuity. View news from FP.
[Alto] SAP Patch Day July 2026: 16 security notes, 4 critical
On July 14, SAP released 16 new security notes, a GitHub advisory, and 3 updated notes: 4 critical, 6 high, 8 medium, and 2 low. The pattern repeats month after month.
And yet, most medium-sized companies continue to patch reactively. An ERP without a disciplined patching cycle is not an asset. It's a liability with compound interest. View newsletter.
[Medium] The European Commission is pressuring: SAP opens its on-premise licensing
Under European regulatory pressure, SAP is loosening its support and licensing policies for on-premise customers. In parallel, it has limited hiring and travel expenses to fund its transformation towards AI.
Two signals worth reading together: SAP is reallocating capital towards AI, and at the same time, losing freedom in its traditional licensing model. For customers with ECC still operational, this opens up negotiation windows that didn't exist a year ago. View coverage.
2. Business Improvement: Processes Preceding AI Agent Governance
[Alto] The AI-powered process optimization market will exceed $113 billion
88% of business leaders say they will increase their investment in AI-powered process intelligence over the next 12 to 18 months. The fundamental shift is methodological.
Manual process mapping is being abandoned in favor of automated process intelligence, which reads transactional systems and reconstructs the actual process, not the one drawn in the manual. The gap between the documented process and the executed process is where most of the inefficiency of a productive company lies. View trends.
[Alto] Operational excellence redefined: no longer efficiency, it's decision speed
The July sectoral analysis marks a conceptual shift. Operational excellence in 2026 will be measured by how quickly a company decides, how it reacts when a supplier stops responding, and whether it truly uses the automation it purchased two years ago.
Directories have stopped accepting promises. This is the uncomfortable metric: how many of the licenses, modules, and automations acquired in the last 24 months are actually in production and with measurable adoption. View report.
[Medium] MIT Technology Review: AI Moves from the Lab to the Production Line
The report published on July 2nd documents AI's transition from the innovation lab to real-world operations. The relevant finding for management is straightforward.
Organizations that capture value are not those that deploy the most models, but those that redesign the process around the model. Without process redesign, AI only accelerates existing chaos. And without an orderly process, any attempt to AI agent government becomes documentary theater. Read article.
[Medium] Between 25% and 55% in productivity gains with operational AI
The range of improvement depends heavily on the industry and the prior level of automation. The strategic read is that AI amplifies existing maturity.
An operation with standardized processes and clean data captures the high end of the range. An operation with dirty data and informal processes captures the low end, or nothing. Data preparation remains the actual bottleneck. See study.
3. Digital Transformation: $3.4 trillion and a 70% that keeps failing
[Alto] The 70% of digital transformation continues to fail in 2026
Despite years of effort and trillions invested, seven out of ten digital transformations fail to meet their objectives. Gartner estimates that only 48% of projects meet or exceed their goals, with a global cost of USD 2.3 trillion annually in failed efforts.
The dominant cause is not technological. McKinsey consistently identifies organizational culture as the main obstacle. Resistance to change, poor change management, and organizational inertia continue to top the list year after year. See predictions. Regarding this point, I wrote before in SAP and digital ergonomics.
[Alto] Global spending on digital transformation will reach USD 3.4 trillion in 2026
The figure is impressive until it's contrasted with the failure rate. Most of that investment doesn't translate into captured value.
The conclusion for a general manager is counterintuitive but clear: the problem isn't the budget, it's execution and adoption. Before approving the next project, it's worth auditing how much of the previous one is actually in use. Statistics.
[Medium] Industry 4.0: A $165 billion market growing at an annual rate of 24%
Growth is driven by smart factories equipped with IoT, AI, and digital twins, capable of adapting to demand in real time. According to PwC, 82% of companies will increase their investment in AI this year.
The stated challenges are consistent across all markets: shortage of specialized talent, cybersecurity, sustainability, and the actual ability to extract value from data. View analysis.
[Medium] Industrial digital maturity is on the rise, but AI adoption remains limited
The CEAM Metal study detects an increase in the level of digital maturity in the industry, with progressive growth especially in medium and large companies. However, the implementation of AI is still limited across the industrial fabric.
The gap between large and medium-sized companies is widening, not closing. Those who do not digitize their core processes in the next 24 months will compete at a structural disadvantage. View prospects.
[Alto] Positive AI returns, but lagging governance structures
A study sponsored by SAP found that companies are seeing positive returns on their AI investments, but their governance structures are not keeping pace.
This gap is the silent risk of 2026: models and agents generating decisions without traceability, without process ownership, and without audit criteria. Today's return can become tomorrow's regulatory liability. It is precisely the void that the AI agent government must close. View coverage.
4. AgTech and Agribusiness Processes
[Alto] Global AgTech investments to exceed USD 30 billion
Q1 2026 saw the highest number of deals in any quarter, 11 deals, but the lowest total capital raised, USD 69 million. This is a sign of a market reopening in seed and Series A stages rather than mega-rounds.
The most funded areas: precision agriculture, AI-powered advisory platforms, smart irrigation, CRISPR gene editing, autonomous equipment, and blockchain traceability. Capital is returning, but with a demand for demonstrable commercial traction. View VC status.
[Alto] Halter raises $220 million at $2 billion valuation
The Series E funding round for livestock infrastructure demonstrates that solutions with clear commercial traction continue to reach unicorn valuations even in a selective capital market.
In parallel, Tropic closed a $105 million Series C in banana and rice genetics, the largest crop genetics round of the period. And AgZen raised $10 million Series B, demonstrating that quantifiable chemical reduction unlocks mainstream capital. The pattern is clear: hard metrics, not agronomic promises. Watch rounds.
[Alto] CampoDigital 2026: Blockchain Digital Passport for 10,000 Producers in Peru
MIDAGRI launched a public-private blockchain traceability initiative covering over 10,000 small-scale fruit and vegetable producers in the Ica and Piura valleys. Each batch receives a digital passport with verifiable data on fertilizer use, harvest date, and organic certifications.
For Latin. See initiative.
[Medium] RFID and Automatic Identification Connect Field, Plant, and Logistics
The integration of RFID, mobile computing, and automatic data capture makes it possible to connect field operations with the plant and export logistics into a single information flow.
The value isn't in the RFID reader but in eliminating manual data entry between steps, which is where inventory errors and unexplained shrinkage occur. In an agribusiness using SAP Business One, this is the integration point with the highest immediate return. See case.
[Medium] Adoption gap: 81% for large farms versus 36% for small producers
Adoption asymmetry is creating two parallel agricultures. For an agribusiness that buys from small producers, this is a concrete operational problem.
The quality of the source data determines the traceability of the entire subsequent process. Investing in digitizing small suppliers is no longer social responsibility. It is chain assurance. View trends.
[Medium] Agroindustrial IoT emerges as a critical attack vector
Sector analysis warns that connected devices in agribusiness will be one of the most critical attack vectors by 2026. Irrigation sensors, cold storage controllers, and packing systems are often deployed without network segmentation or update cycles.
A cold chain attack doesn't steal data. It destroys product. View alert. I wrote about leadership and cybersecurity in Digital Leadership with AI and Cybersecurity.
5. Applied AI in
[Alto] Gartner: $234 billion at risk from agentic AI
On July 1, Gartner reported that up to $234 billion in enterprise application spending is at risk of agency arbitrage between now and 2030. This amounts to about 20% of enterprise SaaS spending.
The logic is simple and disruptive: if an agent can execute the workflow, value migrates from per-user licensing to the orchestration layer. For any IT manager with a portfolio of named-user licenses, this is the most important signal of the month. See statement.
[Alto] The 40% for enterprise applications will have agents by the end of the year
Gartner's projection represents a jump from less than 5% in 2025. The 72% for agent-based AI is already in production.
But the important figure is the other one: 78% of organizations are adopting agents, while 74% are failing to scale them. The disconnect between adoption and value captured is the defining phenomenon of the year, and it explains why the AI agent government it became urgent. See prediction.
[Alto] From demos to workflow replacement: the July change
The pattern among companies that do achieve results is remarkably unglamorous. They map out a messy process, add human review, and test for time saved or errors reduced.
They don't buy narratives. This is the replicable method, and it's the exact opposite of the cross-platform strategy sold at conferences. View summary.
[Alto] AI agent governments become the new cybersecurity
Codenotary launched AgentMon 3, a platform that learns from agent behavior to adapt security policies at runtime. The Cybersecurity Implications of AI Summit on July 9 focused on agentic risk, identity security, and governance strategies.
The operational question for any company is uncomfortable and concrete: if an agent has access credentials to the ERP, who revokes those credentials and under what criteria. The AI agent government This is not an IT project, it's a general management decision. Very brief.
[Medium] APX Coat: agents that execute full credit with audit trail
Abrigo's agentic platform orchestrates and executes complete credit origination workflows: document collection, data review, and exception handling. General availability in Q3 2026.
The stated differentiator is not the agent's capability but the institution-specific audit trails and guardrails. It's proof that the enterprise market is already paying for governance, not just capability. View ad.
Conclusion: The 3 Strategic Signals of the Week
Signal 1: The value of software is shifting from licensing to orchestration. Gartner puts a price on what was already suspected. If an agent executes the workflow, paying per named user no longer makes economic sense. The concrete action for this week: review licensing contracts expiring in the next 18 months and evaluate flexibility clauses. Anyone renewing today for five years under the old model is buying contractual technical debt.
Signal 2: AI adoption has far outpaced its governance, and that gap is the dominant risk. The 78% supports agents, while the 74% is unable to scale them. SAP responds with AI Agent Hub in Q3, and the market responds with products such as AgentMon 3. The AI agent government It should be designed before the third pilot, not after the first incident.
Signal 3: Value capturers map a process, they don't buy a platform. The most consistent finding from July is that companies with measurable results took a single disorganized process, added human review, and measured the results. This stands in direct contrast to the 70% transformations that continue to fail due to cultural—not technological—reasons. For an industrial or agro-industrial company, SAP Document AI for supplier invoices and Process Maps in Business One are two focused, measurable initiatives that deliver a quick return on investment. Starting there is less ambitious and considerably more effective. See also How to optimize SAP Business One with Qlik Sense.
The question this week isn't whether your company will use AI agents. It's already using them, likely without formal authorization. The question is who is responsible when one of them makes a mistake with access to the ERP.
The underlying question remains the same: La Silla Vacía: Who Signs When AI Decides.
