As a child, a girl and her family found themselves living under daily bombardment in the 2006 Lebanon-Israel War. Every day, she waited for something called the United Nations to broker something called a ceasefire.

That little girl was Ola Mohajer. She was a child and a war survivor. And somehow, she knew the UN existed, knew that brokering ceasefires was part of what it did, and believed it would happen. When it finally did, that belief was confirmed. The system worked.

Two decades later, Mohajer looks around at the state of global conflict and wonders: who is brokering ceasefires anymore?

The numbers are grim. According to the 2025 Global Peace Index, global peacefulness has declined for the sixth consecutive year. The report notes 59 active state-based conflicts, the highest number since World War II (and this report predates the current US war with Iran). Conflict-related deaths reached 152,000 in 2024. Perhaps most troubling: the successful resolution of conflicts has fallen to its lowest level in 50 years. Conflicts ending in peace agreements dropped from 23 percent in the 1970s to just four percent in the 2010s.

These are not just statistics. They represent a systemic failure of the tools, institutions, and timelines that the peacebuilding field and multinational organizations like the United Nations have relied on for decades. And they are the reason Mohajer left a 15-year career spanning the Canadian government, the United Nations, and the U.S. Institute of Peace to build Transcend, a startup she describes as the first agentic AI platform for non-kinetic strategy.

What follows is the story of why a war survivor decided the peacebuilding establishment was moving too slowly, what she is building to change that, and the hard questions she is grappling with along the way.Subscribed

Inspired by passion and lived experience, fueled by frustration

Mohajer describes an intersection of her work and life philosophies: “Doing work in the service of humanity drives me in the very core of my being.” She wants to use her time on the planet contributing to tangible impacts for people, especially those who find themselves living through conflicts, like she once did.

This drive was forged by personal experience. Surviving a war as a child gave her a ground-level perspective on international programming that many of her colleagues lack, and shapes her experience working in government and multilateral organizations. She would watch organizations spend millions on initiatives that the local population did not prioritize. And she was frustrated by the slow-moving bureaucracies that seemed to limit impact and spend money unwisely. She would ask her team: if we stopped our work tomorrow and nothing changed, did we actually have any impact? She didn’t want to continue working in organizations where sometimes the honest answer was “no.”

The long lead times for research and lack of planning for sharing findings that were “decision ready” were another big frustration. Taking six months or more to conduct a study, only to have it sit on a desk was a particular thorn-in-the-side for Mohajer.

So when Artificial Intelligence (AI) became more readily available to the masses through the public release of tools like ChatGPT and Claude in late 2022, Mohajer saw an opportunity. She wanted to see how much AI could help with peacebuilding and conflict resolution.

But she hit bureaucratic resistance when she tried to apply AI at her day job. Eventually, she was given permission to explore using AI on the margins of her existing work, with no budget and no allowance for it to distract from her other responsibilities.

For a certain kind of practitioner, such constraints are the tripwire for action. Mohajer began building Transcend in 2023 while still employed. When the momentum of her side project exceeded what the institution could contain in early 2025, she began planning her transition.

The timing, it turns out, was prescient. The organization she was working for was swept up in the collapse of the peacebuilding architecture in the US government in 2025. But Mohajer is clear that she would have left regardless. The frustration was not about one organization. It was about the pace at which the entire field was failing the people it claimed to serve.

A joint staff in a box

So what is Transcend, exactly? Mohajer describes it as a platform that gives policymakers and peacemakers the kind of rapid analytical support that their counterparts in defense and intelligence have long taken for granted.

Mohajer describes the system as operating “like a policy planning cell,” or as others have described it, “a joint staff in a box” and “a dream team of virtual advisors.” The platform uses methods drawn from conflict analysis, political analysis, do-no-harm frameworks, negotiation, and mediation practice to train AI agents that can produce in minutes what traditionally takes weeks or months. Essentially, it automates the research process that uses traditional analytic frameworks and toolkits, reducing the production time of analysis. Then, that automated product is vetted by human experts.

The moment Mohajer knew the technology could work was when Transcend produced a stakeholder matrix. It was the kind of analysis she had spent a month building “by hand” earlier in her career. The system generated it in minutes. She describes her initial reaction as borderline offended.

“But then after I got over my own ego,” she says, “I realized this is amazing. Because the things I really care about are impact. I’m now going to take this, go to the mediation, provide it to the mediator, provide it to stakeholders” and they’re going to have evidence at the ready and always up-to-date.

The key insight here is about what gets automated and what does not. The analytical grunt work, synthesizing data, mapping stakeholders, categorizing actors and interests, is what AI accelerates. The relationship-building, the trust, the phone calls with counterparts in civil society and armed groups, the work of reading a room and knowing when a mediator can push or must wait, all of that remains irreducibly human.

Mohajer spent years doing both. She knows how much of her time went to research that could have gone to relationship-building instead. Transcend’s bet is that freeing practitioners from the analytical bottleneck will make them dramatically more effective at the human work that actually moves the needle toward peace.

The ultimate vision for how Transcend will contribute to impact is a mediator logging in, requesting an analysis on a conflict-affected country like Sudan or Yemen, receiving decision-ready options, and refining them based on their own contextual knowledge. Not waiting six months for an assessment or analysis that will be outdated by the time it gets the mediator’s desk.

Civic voices at scale

Of all the modules Transcend offers, Mohajer is most animated about civic engagement at scale. And this one has a track record that predates AI hype.

In 2020, the Libya Digital Dialogues used technology to enable direct civic participation in the UN-led Libyan Political Dialogue Forum. Under the leadership of Stephanie Williams, the acting head of UNSMIL, online dialogues brought Libyan citizens into the political process in real time.

Participants used smartphones to weigh in on priorities. The questions collected through these dialogues were put to candidates for the Government of National Unity on live television, reaching social media audiences of 1.7 million, roughly a third of the Libyan population.

And it appeared to work. Libya recorded the most significant improvement in overall peacefulness on the Global Peace Index that year, and again the following year.

Mohajer conducted a similar exercise through USIP, orchestrating civic engagement for 600 Sudanese women to speak to the UN Special Envoy to Sudan during the Jeddah mediation talks. The technology available at the time was primitive. Open-ended qualitative questions could not be ingested and analyzed at speed, so she had to design around the limitations, conducting primary and secondary research to identify five priority areas that could be presented as structured choices.

It worked, but it was slow. And the question of attribution remains difficult. Did the engagement directly influence the envoy’s approach? That kind of causal link in mediation cannot be proven the way you can measure a reduction in HIV rates after a five-year health program. But Mohajer argues, and I agree, that dwelling on what you cannot measure is a trap. The peacebuilding field has a persistent bias toward quantifiable outcomes, which often means the most important variables, trust, legitimacy, political will, get undervalued because they resist tidy quantitative metrics.

What excites Mohajer now is the possibility that AI technologies can ingest and make sense of open, unstructured qualitative data from citizens at scale. The technology to do this already exists. The challenge is integration, fine-tuning, and ensuring the engagement is connected to credible, trustworthy processes. As she puts it, the dialogue has to be connected to a person or process in a position of authority who has a reason to listen. And that same authority ideally reports back on how input was used and what change resulted.

Not all dual-use tech faces the same ethical concerns

In every peace tech feature I have written, the dual-use issue comes up. Could the same technology that maps pathways to peace be repurposed for targeting, surveillance, or kinetic operations? It is the question the field has not adequately answered.

Mohajer has a clearer response than most founders I have spoken with, and it is grounded in a technical distinction that is important to understand.

Transcend is what the industry defines as being in an AI vertical. It is agentic AI, purpose-built for a specific domain—peace, security, and strategic risk—using methods and training data drawn from that domain. The agents are trained on mediation frameworks, conflict analysis methods, the Berghof Foundation’s mediation resources, and similar bodies of practice-specific knowledge. This is fundamentally different from a foundational model like those built by Anthropic or OpenAI, which are general-purpose systems designed to be applied across every domain. Agentic AI runs on top of foundational models, the way a web application runs on the internet. But the web application layer is purpose-built.

In this way, foundational models are like the internet, and AI verticals are like websites or apps you build on the internet. Yes, the AI vertical could be changed, like apps can be changed, but it would take a lot of effort—recoding, redesigning, etc. If your app is about birds, and a customer wants an app about cars, you could repurpose your bird app to be about cars; or, the customer could just go to another pre-existing app about cars. It’s easy to see how inefficient it would be to do this, especially when car apps (or in Transcend’s case, defense tech) already exist.

To repurpose Transcend’s agents for military targeting or intelligence operations, she argues, you would need to retrain them with entirely different methods, data, and instructions. There are plenty of companies already building agentic AI within kinetic domains. Transcend is not one of them, and its architecture would not be a natural starting point for such purposes.

Much of the dual-use anxiety in peace tech is directed at foundational models, the internet layer in Mohajer’s framing, rather than the applications built on top of them. Foundational model providers like Anthropic face hard questions about how their technology might be used because their systems are, by design, general-purpose. This is not the same as a vertical application built for a specific domain. While not immune to misuse, but the pathway to misuse requires more deliberate effort.

That said, architecture is not a permanent shield. As I’ve covered in past features, founders leave; companies get acquired; business models shift under investor pressure. Which brings us to the question of who gets to shape the company’s future.

Who sits at your cap table matters

Mohajer is candid about the governance challenges facing mission-driven startups. She has clearly spent time thinking about what happens when the founder’s values are no longer the only thing standing between the company and a bad decision that could result in harm.

Her approach starts with decisions on who invests in Transcend. She describes herself as interested in double bottom line investors: those who want to see both profit and social impact. She knows that narrows the field of potential investors. Many venture capitalists are not interested in peace and security. But she sees that self-selection as a feature, not a limitation.

She is also clear-eyed about the limits of good intentions. Investors with board seats do not just bring capital. They bring pressure, expectations, and the power to shape decisions. Mohajer tells the story of a managing partner at a VC firm who loved Transcend’s mission and product. They entered due diligence. Then a partner at the firm who was focused on financial returns decided it was not the right fit and they parted ways. Transcend’s human-at-the-center approach, which includes bespoke guardrails and real human-expertise informing analysis and decisions, has a cost and those small but critical costs were deemed to erode the extremely high returns that fund was seeking. In Mohajer’s framing, that is self-selection working as it should. Human expertise is the moat for Transcend. It is also the very thing that wins the trust of clients in these high-stakes domains.

On governance structures more broadly, she argues that these cannot be static. A fast-growing company needs dynamic governance that evolves with the organization. She compares it to raising children: the rules for a toddler do not work for a teenager. What matters is that governance is fit for the moment and purpose of what the technology is being used for.

These seem like the right instincts. Whether they survive contact with the kind of capital pressure that has distorted companies far larger than Transcend remains to be seen. The history of mission-driven technology companies is littered with cautionary tales of founders who believed their values would be enough, but were acquired by larger firms who had a different set of priorities.

Given Transcend’s tech stack and purpose, as well as Mohajer’s clear-eyed and value-driven leadership, it could be a great success story of avoiding these industry pitfalls.

The 20-year question

I asked Mohajer what wild success looks like. Her answer was not about revenue or market share.

“It’s that we have far less violent conflict than we’re seeing today,” she says. She points to the sheer volume and protractedness of current conflicts. The trend lines are moving in the wrong direction by nearly every measure. She suggests that in 20 years, violent conflict could be obsolete. Then she catches herself: that might be sci-fi. But in AI terms, 20 years is a millennia.

The more grounded version of her vision is a world where Transcend’s software is used by governments and companies globally as a trusted technology partner for identifying options that maximize outcomes at geopolitical and economic levels. Less war. Greater prosperity. And prosperity that is equitable, though she notes that the equity part is left to governments that inevitably make political decisions.

The ambition is large. And it rests on a set of assumptions that the field has not yet fully validated: that AI-assisted analysis can meaningfully compress the timeline to peace; that civic engagement at scale can be connected to processes with the authority and will to act on it; that a mission-driven startup can navigate the pressures of venture capital without losing its peace tech roots.

This is the latest in a series of features profiling companies and leaders in the peace economy. If you are building companies or investing in this space and want to be part of the conversation, I’d welcome hearing from you. Book a call with me here.