AI fatigue is real, but it is selective. Buyers are not rejecting AI. They are rejecting outreach that reads as mass-produced and irrelevant. In B2B outbound for Web3 and FinTech, our position is simple: use AI for research, list-building and first drafts, and keep a named person accountable for the message, the claim and the reply.
This guide covers what the research says, where we draw the line, and how to test whether AI is costing you replies.
_____________________________________________________________
Quick Answers – Jump to Section
- What does the research say about AI fatigue?
- Why does AI fatigue show up quickly in outbound?
- Where does AI help in outbound, and where does it hurt?
- How do we split the work between AI and people?
- What about AI phone agents and voice?
- What should you check before an AI-assisted message goes out?
- How do you measure whether AI use is hurting replies?
- What should Web3 and FinTech teams do first?
- Frequently asked questions
_____________________________________________________________
What does the research say about AI fatigue?
The public is wary of AI, but willing to use it for some tasks. A Pew Research Center survey of 5,023 US adults, fielded in June 2025, found that 50% are more concerned than excited about the increased use of AI in daily life. That is up from 37% in 2021, and only 10% said they are more excited than concerned. The remaining 38% are equally excited and concerned. You can read the figures in Pew’s report on how Americans view AI.
The same survey shows the selectivity. Nearly three-quarters, 73%, would let AI assist them at least a little with day-to-day tasks, but only 13% would let it help a lot. Americans were also more open to AI for data-heavy work, such as weather forecasting and developing medicines, than for personal subjects. Half of respondents, 50%, said AI will worsen people’s ability to form meaningful relationships, against 5% who said it will improve it.
There are limits to this evidence. It is a general-public survey of US adults, so it does not tell us how buyers at banks, funds or protocols react to AI-written outreach. Treat it as a direction, not a measurement of your market. For the B2B side, see the Gartner figures in the next section.
_____________________________________________________________
Why does AI fatigue show up quickly in outbound?
Outbound is unsolicited, so the reader decides within seconds whether a person wrote the message and whether it is relevant. Gartner’s 2025 survey of 632 B2B buyers found that 73% actively avoid suppliers who send irrelevant outreach. Gartner’s own analyst put it bluntly: bad prospecting actively damages relationships. The press release is on Gartner’s site.
Notice what that figure measures. It is about irrelevant outreach, not about AI. AI makes irrelevant outreach cheaper to produce, so buyers see more of it. That is the real source of fatigue.
Buyers in banks, funds and protocols face a second problem. They read claims about compliance, custody and yield closely, because a wrong claim creates risk for them. A message that sounds confident but gets a regulatory detail wrong does more damage than a message that says less.
_____________________________________________________________
Where does AI help in outbound, and where does it hurt?
AI helps with work that is largely data, and it hurts when it speaks for you. That mirrors the Pew finding that people accept AI for data-heavy tasks more readily than for personal ones. Outbound has both kinds of task, and we treat them differently.
| Task | AI role | Person’s role |
|---|---|---|
| Account research | Pulls signals and summarises a company’s recent news | Checks that the signals are true and relevant |
| List building and segmenting | Sorts accounts by fit and trigger | Sets the criteria and removes bad fits |
| First drafts and variants | Produces options to react to | Writes the first line and the ask |
| Claims about regulation, licences or yield | None | Owns every word |
| Replies | Suggests a draft | Reads, decides and sends |
| Deciding not to send | None | Makes the call |
The pattern is consistent. AI prepares, and a person speaks. If a task needs judgement about a relationship or a regulated claim, a person does it.
Our cold email templates for blockchain infrastructure buyers follow the same rule: the template is a starting point, and the first line is always written for the recipient.
_____________________________________________________________
How do we split the work between AI and people?
We let AI do the research and the heavy lifting, and a person writes and approves everything that reaches a buyer. In practice the workflow has five steps.
- Research. AI helps gather and summarise signals about the account, such as a funding round, a licence application or a hiring pattern.
- First line and ask. A person writes these, because they carry the relationship.
- Variants. AI can suggest alternatives for subject lines and body copy, which a person edits or rejects.
- Review. A person checks every message against the checklist in the next section.
- Reply. A person reads and answers every reply.
This is also where AI SDR rollouts tend to go wrong. We cover the failure patterns in why 98% of AI SDR implementations fail in FinTech.
_____________________________________________________________
What about AI phone agents and voice?
Voice raises the stakes, because people react more strongly to a call than to an email. A person who answers a call and finds an AI on the line may feel tricked, and that feeling attaches to your brand. Our rule is outbound first, answering second, and a person owns the script and the escalation path.
Start with disclosure. Tell the caller at the start of the call that they are speaking to an AI. Rules on this differ by country and are changing, so check the current position with legal counsel before you launch.
Then keep the first use narrow. Use voice for tasks where the caller expects automation, such as booking a slot or confirming details, before you use it for cold outreach. Make sure a person can take over quickly, and that the caller can ask for one at any point.
_____________________________________________________________
What should you check before an AI-assisted message goes out?
Use a short checklist, and do not send a message that fails any item. We apply six checks.
- Is every claim true? Check licences, regulations, numbers and names against a source.
- Is it specific to the recipient? If the message could go to 500 other people unchanged, rewrite it.
- Does it give a real reason for contact? A trigger, such as a launch, a filing or a hire, beats a general compliment.
- Is it compliant? Avoid guarantees and promises about returns, and respect consent and data rules. Our guide to GDPR and cold emails covers the basics.
- Can the sender defend each line? If the person whose name is on it would hesitate to repeat a line on a call, remove it.
- Who answers the reply? Name the person and the time limit before you send.
The checklist is short on purpose. A long one gets skipped.
_____________________________________________________________
How do you measure whether AI use is hurting replies?
Run a controlled test on a single segment, and compare three versions of the message. Version one is written by a person. Version two is drafted by AI and edited by a person. Version three is AI-only. Send each to a similar group of accounts over the same period.
Track these measures for each version:
- Reply rate
- Positive reply rate
- Meetings booked
- Unsubscribes and complaints
- Spam placement, if your tooling shows it
Keep the test running for several weeks, because B2B replies arrive slowly. Do not draw conclusions from a few days of data. Change one variable at a time, because if version two also has a different subject line, you will not know what moved the result. Record each send in a simple sheet with the segment, the dates, the volumes and the outcomes.
For a wider framework on proving outbound results, see how to measure ROI from LinkedIn outbound for FinTech and Web3 companies.
_____________________________________________________________
What should Web3 and FinTech teams do first?
Start with an audit of where AI already touches your outbound, then add one human review rule. You do not need new tools to do this. It is common for a first audit to find tasks that nobody decided to automate, such as follow-ups and reply drafts that went live because a tool made them easy. Write the rule down and share it with everyone who sends on your behalf, including contractors and agencies.
- List every task AI touches. Include research, drafting, personalisation, follow-ups and replies.
- Mark the tasks that speak for you. These are the ones where a person must review the output.
- Add one rule. For example: no message goes out without a named person approving the first line and every claim.
- Run the three-version test from the previous section on one segment.
- Review the results after several weeks, then adjust the split.
If you are building an outbound programme from scratch, the appointment setting playbook for Web3 and FinTech B2B sales shows where each step fits.
_____________________________________________________________
Frequently asked questions
Is AI fatigue real among B2B buyers?
The evidence points to fatigue with irrelevant outreach rather than with AI itself. Gartner found that 73% of B2B buyers avoid suppliers who send irrelevant outreach, and Pew found that half of US adults are more concerned than excited about AI. Neither survey measures reactions to AI-written outbound at banks, funds or protocols, so test your own market.
Should we stop using AI in outbound?
No. Use it for research, list-building and drafts, and keep a person responsible for the first line, the claims and the replies.
Can buyers tell when a message was written by AI?
Sometimes, and many do not mind AI use as long as the message is relevant and accurate. The risk is a message that reads as generic. Check each message for specifics that only a person who researched the account would include.
Should we tell prospects that AI helped write a message?
For calls, tell the person at the start. For written messages, we require that a named person approves and is accountable for the content.
How many messages should a person review?
All of them. If volume makes that impossible, reduce volume until a person can read every message before it goes out.
How long should an outbound test run?
Several weeks at least, because B2B replies arrive slowly. Compare the same segment over the same period, and do not change the messages mid-test.
We run outbound for Web3 and FinTech companies selling to banks, funds and protocols.
_________________________________________________________________
How visible are you on ChatGPT?
We’re currently testing our new AI Visibility Auditor to track brand citations across FOUR AI search engines.
We ask 120 questions to find out whether your business is named, who gets named instead, and on which sites it found that information.
Follow the report to see where AI engines place your brand today, and the steps to become the brand they recommend first.
We’re opening up a free beta group for a small group of founders before our Black Friday £499/month launch.
If you’d like us to test your site and join the beta, share below how many times you think your brand will be cited from 120 questions?
Want to know how we can guarantee a mighty boost to your traffic, rank, reputation and authority in you niche?
Tap here to chat to me and I’ll show you how we make it happen.
If you’ve enjoyed reading today’s blog, please share our blog link below.
Do you have a blog on business and marketing that you’d like to share on influxjuice.com/blog? Contact me at rob@influxjuice.com.
Latest Blogs
- AI Fatigue in B2B Outbound: Where Web3 and FinTech Buyers Draw the Line
- ElevenLabs Review (2026): ElevenAgents, ElevenAPI and ElevenCreative Explained
- Emergent Agentic Vibe Coding Platform: The Complete Guide to Building Apps Without Code
- How Stablecoin B2B Projects Are Using Thought Leadership to Build Confidence With Enterprise Buyers in Europe
- What’s Making AI Impact Difficult to Measure for Marketers?
