42 MAY-JUNE 2026 | THE FOREIGN SERVICE JOURNAL local newspaper”). If, instead, that officer cultivates a relationship with a public official or mining executive and gathers original insights conveyed in the report, that is good. But how good? One might imagine the creation of a grading system that rates any officer’s submission. If I do a report on a young, ambitious mining ministry official for whom there is little public source reporting, AI could be taught to rate that highly. And even more significantly, as AI synthesizes public sources and human intelligence from post that is then referenced by yet other reports on, say, the mining sector in Latin America, a grade for relevance based on number of citations could be added as well. Such a system is clearly possible with today’s technology. This sort of grading would encourage officers to make longer-term investments in relationship-building that pay off for U.S. national interests well beyond the annual evaluation or departure from post. Imagine a scenario in which a relationship with a charismatic young labor leader like Luis Inácio Lula da Silva was developed and reported on—and paid off in dividends for years until he eventually became president of Brazil. AI could continue to flag the relevance of that initial relationship and ensure credits through a citations index for the officers who maintain the relationship as Lula grows in prominence. The goal is to motivate reporting officers to get outside embassy walls and develop a wide array of relationships that are of national interest to the U.S., both in the short term (when, for example, a project calls for engaging them to block Chinese port investment) and in the long term (when a young private sector leader has presidential aspirations). A CRM platform with a case management module would allow ad hoc teams to address a problem with the ability to document their work. Commerce’s well-established Salesforce CRM case module, used to capture hundreds of billions of dollars of U.S. business wins globally, serves as a useful prototype. More important than the technology is senior leadership’s insistence that measurable goals for outcomes be defined, developed, and captured on one platform, across all relevant agencies. And that platform is logically where all profiles of individuals, companies, NGOs, and government entities are maintained and linked as needed to a case’s execution. Commerce currently insists that its global workforce capture such information on its Salesforce CRM in order for anyone to get credit for their work. Ambassadors can make the same demand, and OMB can force agencies to use this common platform when at post. What Could Go Wrong? The State Department may not be ready. Changing the culture of reporting officers at State will not be easy. But after talking with multiple colleagues there, I firmly believe there is a newfound awareness that adapting or dying are today’s options. State’s senior-most leadership is telling ambassadors that there needs to be a change in business as usual and that they will themselves be held accountable for execution. And the report card approach—based on developing and documenting more relationships of interest, making original contributions, and finding relevance through citations—may focus the mind quickly if it is factored into evaluations. AI may not be ready. If AI can’t reliably tag and structure the less-ordered intake of knowledge from, say, voice-activated dialogue, users will not embrace it. Early AI implementation in other U.S. government agencies has been met with considerable skepticism. And piloting an AI-dominant solution at a few posts will flag issues. Other agencies must be on board. As the president’s representative at post, the ambassador will need to insist on participation—no more working in silos—and be willing to send agency reps home if they aren’t team players. When used widely, AI will become a more valuable resource for all agencies at post because its reports draw from a deeper well of knowledge. Securing knowledge is tricky. It is almost certainly the case that profile management and methods for exerting leverage to win a mining concession need to be classified. That, in turn, may constrain some of the more innovative private sector solutions. And need-to-know protocols will have to be designed and implemented so that knowledge is compartmentalized and protected, but also available to those who need it. There may be pressure to have parallel systems, classified and unclassified, which will almost surely tank user adoption or, worse, result in leaks of knowledge that should be protected. This issue needs to be hashed out thoroughly, but a simple if draconian approach would be to insist that the entire platform AI cannot replace human intelligence gathering, the relationships that embassy officials develop.
RkJQdWJsaXNoZXIy ODIyMDU=