Forrester published its 2027 B2B go-to-market predictions on 21 October, under a headline built for board decks. AI agents, it says, will influence 75% of B2B purchase decisions.
That figure is a prediction, and Forrester words it with some care. Its predictions page says agentic buying “will play a role in 75% of B2B purchases in 2027”. Playing a role is a wide net. It covers a buyer asking an assistant for a list of suppliers as much as an agent running a contract-renewal analysis.
The prediction that should worry a marketing team sits further down the release. Forrester expects two-thirds of providers to lose deals because buyers cannot verify their claims. That one is about your content, and you can act on it this quarter.
Key takeaways
- Forrester predicts agentic buying will play a role in 75% of B2B purchases in 2027.
- It also predicts two-thirds of providers will lose deals because buyers cannot verify their claims.
- Agents favour proof they can check, such as case studies with figures, public pricing and third-party reviews.
- Figures that differ between your site, your deck and your review profiles read as weak claims.
What Forrester actually predicts
The release lists three predictions.
- Buyers will double their human interactions to validate AI-generated recommendations. AI cannot fully judge fit, feasibility, trust or accountability, so buyers will check with providers, analysts, references and partners.
- Two-thirds of providers will lose deals because buyers cannot verify claims. Buyers already use AI-assisted search to synthesise and fact-check vendor claims.
- Most CMOs will cut programmes or accept reduced capacity after overestimating AI gains. Some will eliminate as many as 25% of roles for productivity gains that fall short.
The release also draws a limit. Truly autonomous buying will stay concentrated in high-volume, low-risk and repeatable purchases. Your enterprise deal will still be signed by people. The agent works earlier, when it researches suppliers and shapes the shortlist.
In a companion post, analysts Karen Tran and Anthony McPartlin set the horizon at the end of 2027. AI tools, they write, “rapidly synthesize provider claims and check facts, exposing hollow or inconsistent sales and marketing statements.” Claims must be backed by “proof points that AI tools can index and validate.”
The full report is reserved for Forrester clients, so we have not seen the method behind the 75%. We read it as a direction of travel.
What a buying agent does with your website
Picture the brief a procurement lead gives an assistant. Find five providers for a company our size in the UK, compare their pricing, check their security credentials and summarise what customers say.
The agent reads public pages, documentation and review sites, then lines the claims up. A claim with a figure, a date and a named customer behind it can be repeated with confidence. “Industry-leading” gives it nothing to extract, and that line of the comparison stays empty.
An empty cell next to a competitor’s filled one is how you drop off a shortlist without ever hearing about the deal.
The proof an agent can read and believe
Five types of content carry weight in this kind of evaluation.
- Case studies with figures. A named customer, or a precise description if the name is confidential, the starting point, the period and the result in numbers. “Drove significant growth” gives an agent nothing to compare.
- Public pricing, or at least public pricing logic. Starting prices and what pushes them up. “Contact sales” leaves the agent with an empty cell.
- Documentation and security pages. Integrations, certifications, data residency. Many vendors still hide these in PDFs sent after a first call.
- Honest comparisons. A page that says where a competitor is the better fit is rare, and it gives the agent a reason to trust you on the cases where you win.
- Third-party evidence. Review platforms, analyst mentions, partner listings and customer references. Forrester expects buyers to lean harder on these to validate what AI tells them.
Some content carries little weight. Gated PDFs an agent cannot open, logo walls without stories, superlatives without a figure, and numbers that change from one page to the next.
Consistency is the part teams underestimate
Most B2B companies carry several versions of the truth. Say the website claims 300 customers, the sales deck says 250 and the review profile has not been touched in two years. A human reader shrugs. An agent comparing sources sees a claim it cannot confirm, which is the “hollow or inconsistent” statement Forrester describes.
The fix is unglamorous. Pick the ten claims you rely on most, agree on the true current figure for each, and update every place it appears, including the profiles your team does not manage day to day.
B2B AND ABM
Make your proof readable by people and agents
We audit what AI assistants say about you, rebuild the case studies and pages they rely on, and tie that proof to your target accounts.
”Our deals are too complex for an agent”
This is the objection we hear first, and Forrester half agrees with it. Autonomous buying stays in simple, repeatable purchases, and large deals will involve more human conversations.
Look at who those humans are, though. Analysts, references and partners are all sources of third-party proof, and the agent often brought the buyer to them. In a complex sale the verification happens twice, first by the machine, then by the people it pointed to.
If your partners quote different figures from yours, that second check fails too. Our partner marketing guide covers how to build joint proof with them.
A 90-day plan for a B2B marketing team
Weeks 1 and 2. Ask two or three AI assistants to compare you with your two closest competitors for a typical brief. Save the answers and note what they get wrong or leave out.
Weeks 3 and 4. List your ten main claims, the proof behind each one and every place it appears. Rewrite or delete the claims you cannot prove.
Weeks 5 to 8. Rebuild three case studies around figures and dates, and ungate them. Publish a pricing page, even if it only gives ranges. Move security details from PDFs to indexable pages.
Weeks 9 to 12. Run the consistency pass across review platforms, partner listings, LinkedIn and the sales deck. Then rerun the first test with the same prompts.
How pages get picked up by AI answers is the subject of our LLMO guide. This plan deals with what those answers find once they reach you.
An agent cannot repeat a claim you never proved. It leaves the cell empty, and a competitor fills it. MyDigipal
Where we would start
If budget forces a choice, start with case studies. Most B2B companies already have the raw material, buried in account managers’ notes and quarterly reviews. Written up with figures, the same proof serves the agent’s comparison, the sales deck and the reference call.
Then connect that proof to the accounts that matter. The intent signals you already track show which accounts are researching your category, and that is where a missing figure costs the most.
Forrester’s CMO prediction carries a warning. Teams that cut people expecting AI to absorb the work will lack hands for this job, which is slow and needs someone chasing customers for permission to publish results. We would protect that capacity first.
Sources: Forrester press release, 21 October 2026 - Forrester blog on the 2027 predictions - Forrester 2027 predictions guide
