For most of human history, trust was closely tied to presence. Important decisions happened in rooms. Agreements were negotiated face-to-face. Documents were reviewed in the company of other people. Signatures were observed. Questions could be asked and answered in real time. The process was not perfect, but it provided something organizations rarely needed to think about explicitly: verification.
Presence acted as evidence.
When people gathered to execute an agreement, they were not simply exchanging information. They were establishing context. They could observe one another, assess intent, and verify participation. The value of physical presence was never the room itself. The value was the confidence created by the room.
As business became increasingly digital, organizations began removing that requirement. Email replaced physical correspondence. Video conferencing reduced travel. Digital agreements accelerated transactions. Entire industries embraced the idea that important business could occur without requiring participants to occupy the same space.
The transition was rational and largely successful. Digital transformation reduced costs, increased speed, and removed unnecessary friction from commercial activity. Agreements could be executed across cities, countries, and continents without requiring a single flight or conference room reservation. Businesses became more efficient because presence was no longer required.
What many organizations did not fully appreciate was that they were removing more than geography.
They were also removing a verification mechanism.
The systems that emerged during this period inherited assumptions from the physical world. An email was assumed to originate from the person whose name appeared in the sender field. A face on a video call was assumed to belong to the participant speaking. A voice was assumed to belong to the individual everyone believed it belonged to. The signature itself did not need to carry the full burden of trust because identity seemed relatively certain.
This assumption made perfect sense when the technologies involved were difficult to imitate. The surrounding signals provided enough confidence that organizations could focus primarily on efficiency rather than verification. Digital signatures, audit logs, and workflow systems were built within this environment. They were designed to record transactions, not necessarily to defend them against increasingly sophisticated identity challenges.
The environment has changed. Fraud and the manipulation of AI are on the rise.
Artificial intelligence has dramatically altered the economics of deception. Faces can now be generated convincingly. Voices can be cloned from short audio samples. Writing styles can be replicated. Video calls can be manipulated. The signals organizations once relied upon to establish identity are becoming increasingly unreliable.
This shift is not hypothetical.
Deloitte projects that generative-AI-enabled fraud losses could reach approximately US$40 billion annually by 2027, compared with roughly US$12.3 billion in 2023.¹ The projection reflects a broader concern that artificial intelligence is accelerating the ability of attackers to impersonate trusted individuals, automate fraud, and exploit systems that rely heavily on assumptions about identity.
The implications are already visible in real-world incidents.
In 2024, an employee at global engineering firm Arup participated in what appeared to be a routine video conference involving several colleagues. During the meeting, instructions were provided and approximately US$25 million was transferred. Subsequent investigations determined that every participant appearing on the call had been generated using deepfake technology. The meeting itself occurred. The instructions were delivered. The participants appeared authentic. The underlying identities, however, were entirely fabricated.² ³
The significance of this incident extends beyond the financial loss. The event demonstrates that many of the signals organizations historically relied upon to establish trust can no longer be assumed to provide reliable verification. Participation is not proof of identity. Appearance is not proof of authenticity. Presence, at least in digital form, is no longer necessarily evidence.
Research suggests that this challenge is likely to intensify.
A 2025 study conducted by iProov found that only 0.1% of participants successfully identified all examples of real and synthetic media presented to them. Even when instructed to actively look for deepfakes, the overwhelming majority struggled to distinguish authentic content from AI-generated content.⁴ The study highlights an uncomfortable reality: human observation, which served as a practical verification mechanism for centuries, is becoming less effective against technologies specifically designed to deceive it.
At the same time, broader fraud trends continue moving in the wrong direction. According to the U.S. Federal Trade Commission, consumers reported approximately US$12.5 billion in fraud losses during 2024, representing a significant increase over previous years.⁵ While not all fraud involves artificial intelligence, the data illustrates a broader pattern: trust-based systems are facing increasing pressure from more sophisticated methods of deception.
Many organizations are responding by focusing on records. They want stronger audit trails, additional timestamps, enhanced monitoring, and more comprehensive documentation. These measures are useful and often necessary. However, they do not necessarily address the underlying issue.
A record tells us that something happened.
Verification tells us who made it happen.
The distinction is increasingly important because many organizations continue to treat these concepts as interchangeable. They are not. A timestamp can establish when an event occurred. An audit log can establish the sequence of actions. A workflow can establish how a process unfolded. None of these mechanisms necessarily establish identity.
This is where the most important shift is occurring.
The problem is not that digital agreements stopped working.
The problem is that many organizations continue to rely on trust signals that were designed for a different environment.
The problem is not distance.
The problem is certainty.
For years, the market has discussed digital transformation primarily in terms of convenience, speed, and efficiency. Those benefits remain important. However, the next phase of digital transformation is increasingly about confidence. Organizations must determine not only whether a transaction occurred, but whether they can demonstrate who participated, how identity was established, and why the transaction should be trusted.
This challenge is creating the conditions for a new category.
Traditional discussions focus on signatures, workflows, and records. Increasingly, the market is moving toward a broader question: how do organizations establish trust in environments where identity can no longer be assumed?
The answer is unlikely to be found in additional records alone.
It will increasingly be found in evidence.
Evidence-grade agreements represent a different way of thinking about trust. Rather than focusing exclusively on recording transactions, evidence-grade agreements focus on establishing confidence in the individuals participating in those transactions. They recognize that trust is no longer simply a function of process. Trust is becoming a function of verification.
This distinction may prove increasingly important as artificial intelligence continues to improve.
The challenge facing organizations is not whether digital transactions will continue. They will.
The challenge is determining which trust assumptions remain valid and which require rethinking.
For decades, digital transformation removed the need for physical presence. That transition delivered enormous value. Artificial intelligence is now forcing organizations to confront a more difficult question: if presence is no longer reliable proof of identity, what replaces it?
The answer to that question will likely define the future of digital trust.
Sources
- Deloitte Center for Financial Services, “Generative AI is expected to magnify the risk of deepfakes and other fraud in banking,” 2024.
- Financial Times, “Arup lost $25mn in Hong Kong deepfake video conference scam,” 2024.
- CNN, “British engineering giant Arup revealed as $25 million deepfake scam victim,” 2024.
- iProov, “Study Reveals Deepfake Blindspot,” February 2025.
- U.S. Federal Trade Commission, Consumer Sentinel Network Data Book 2024, released March 2025.