The Future MEL System Is Conversational
How AI can make reflection as natural as dialogue
Evaluation did not begin as a learning enterprise. It began as an experiment in proof, rooted in education and later scaled to government programs that needed to demonstrate accountability. The early focus was measurement and verification: did the intervention work, and was it worth the cost?
By the 1970s, theorists like Robert Stake and Michael Scriven questioned that narrow view. They argued that understanding a program’s context and
the perspectives of those involved mattered as much as outcomes. In the years that followed, Michael Patton’s Utilization-Focused Evaluation brought learning and use to the center. Participatory and developmental approaches built on this momentum, expanding evaluation beyond judgment toward adaptation and collective sense-making.
That learning tradition is gaining new momentum. Across the field, evaluators are working to rebalance accountability with learning. The UNDP’s M&E Sandbox asks what it would take to center learning in monitoring practice. The Water for Women Fund has built triple-loop learning directly into its MEL system, asking “Are we doing things right? Are we doing the right things? Why are we doing them at all?” The International NGO Training and Research Centre (INTRAC), a UK-based organization that helps civil society groups strengthen their MEL systems, has been promoting “collective learning” spaces where funders and implementing partners learn together. These examples reflect a larger shift. Evaluation is moving from reporting to reflection, from data collection to dialogue.
Artificial intelligence has a role to play in that evolution. Recent discussions at the American Evaluation Association and in the Learning for Sustainability community suggest cautious optimism. Generative AI can summarize large volumes of qualitative data, surface patterns, and prompt evaluators to consider alternative interpretations. Used well, it supports learning by freeing time for sense-making rather than data handling. Used poorly, it risks flattening judgment into automation.
At Anthralytic, this distinction guides our design. The Evaluability Self-Assessment app helps teams notice readiness gaps and surface assumptions before an evaluation begins. The Theory of Change Wizard works the same way. As users map outcomes and causal links, the system prompts questions such as “What evidence supports this?” or “Who benefits here?” The tools do not replace facilitation; they structure reflection.
This is what a conversational MEL system looks like in practice:
Data systems that prompt rather than store.
Interfaces that make reflection routine rather than optional.
AI that expands human learning rather than imitates it.
Technology is not rewriting evaluation. It is catching up with it. The field’s evolution from proof to learning has always been conversational at its core—people exchanging evidence, stories, and meaning. AI can help sustain that conversation between reports and reflection sessions, making learning an everyday act instead of an annual event.
The future of MEL is not more data. It is better dialogue.
Anthralytic helps social impact organizations sharpen strategy, evaluate results, and build systems that learn. Our hybrid consulting and AI tools make reflection practical for teams that are short on time but serious about impact.
