A client rarely leaves on the day the relationship breaks. The fracture arrives earlier and in smaller pieces: a sponsor stops joining the Tuesday call; a formerly brisk email becomes carefully polite; a promised introduction does not happen. By renewal time, the story feels sudden only because nobody assembled the clues. Kaizan, a London startup founded by Glen Calvert and Pravin Paratey in 2022, is trying to make those clues legible while someone can still do something about them.
Its platform pulls from meetings, email, chat, documents and account systems, then builds a running picture of each client. It can draft recaps, flag a thinning stakeholder map, prepare a brief, update a CRM or suggest an expansion play. Humans approve the work that matters. The larger pitch is that an agency, consultancy or customer-success organization can turn private account-manager intuition into shared institutional memory.
The gap between a signature and a renewal
Enterprise software has thoroughly instrumented the journey to a sale. Leads move through stages, opportunities acquire probabilities and contracts land in systems of record. The years after the signature are untidier. Much of the truth lives in conversation, and much of the operating context lives in the head of the person closest to the account.
Calvert's founder letter makes the distinction plainly: the CRM was built to track the trip to a signature, not the long delivery that follows it. That insight is Kaizan's wedge. The company does not claim to replace Salesforce, HubSpot or Pipedrive. It describes those tools as the record and Kaizan as the system of work alongside them.
The signal was always there, trapped in conversations no software could read.Glen Calvert, Kaizan co-founder and CEO
The line explains both the opportunity and the risk. Language models can now summarize a call, detect topics and compare patterns across thousands of interactions. But a confident summary is not necessarily a faithful reading of a relationship. Kaizan's answer is evidence: a risk or sentiment judgment is supposed to arrive with the underlying interaction attached, so an account manager can inspect the reasoning instead of receiving an unexplained red light.
Four letters around a messy relationship
In June, Kaizan introduced CARE, its framework for relationship intelligence. The current product describes the letters as Client sentiment, Account activity, Relationship coverage and Engagement growth; its FAQ uses the nearby labels Client Satisfaction, Activity, Relationship and Expansion. The language varies slightly, but the underlying questions are recognizable: How does the client feel? Are the right people talking often enough? Is the relationship broad and senior enough? Is there evidence the work can grow?
A useful health model does more than compress uncertainty into a number. It makes disagreement productive. If the score says a relationship weakened because a senior buyer dropped from the meeting cadence, the team can challenge the premise or act on it. Kaizan says CARE is calibrated to each customer's own retained and lost work rather than an industry-wide average, with outcomes feeding the model each week.
That tenant-specific approach matters. A warm relationship at a 20-person design studio does not look like one at a multinational consultancy. It also creates a practical test for buyers: ask how quickly the model becomes useful when a firm's own history is thin, inconsistent or stored in inaccessible places.
Where Kaizan sits in a crowded stack
The adjacent names are familiar. Pendo combines software-usage analytics, product feedback and in-app guidance. Whatfix focuses on digital adoption, placing guidance and self-service help over applications while measuring friction. Gainsight offers a broad customer-success platform with health scores, renewal workflows, product usage, sentiment and AI agents. Its current product already analyzes communication signals, which makes it the closest large incumbent in this frame.
Kaizan's distinction is a matter of starting point and customer. It begins with client-service communications and the work surrounding an account, particularly inside agencies, consultancies and professional-services firms. The companies overlap; no honest market map makes them neat substitutes. A large SaaS business might use product telemetry from Pendo, in-app guidance from Whatfix and customer workflows from Gainsight. Kaizan is betting that a service-led firm will prefer a system trained on conversations, deliverables and stakeholder dynamics.
Money for the unglamorous middle
In May 2026, Pembroke VCT led a funding round of up to £2.5 million, investing £1.5 million alongside Velocity Capital and industry angels. Pembroke says the money will support go-to-market work, international expansion and continued development of Kaizan's AI capabilities and CARE model. The investor lists customers including Gravity Global, Tradedoubler and The Gap Partnership, and says Kaizan handles communications in more than 30 languages.
Funding and operating figures are reported by lead investor Pembroke VCT.
Kaizan publishes stronger performance numbers of its own. Its FAQ says account managers save an average 3.2 hours a week on reporting and follow-up work, and that QBR preparation falls from four to six hours per account to between 30 and 60 minutes. Pembroke cites case studies showing more than five hours saved and average revenue per client rising by over 20 percent. These are company and investor claims, not independent benchmarks, and prospective buyers should ask for comparable customers, starting baselines and measurement methods.
The part worth stealing now
Even without Kaizan, an account team can borrow the operating idea. Stop treating health as a color chosen before a meeting. Separate the relationship into observable dimensions. Require evidence for every risk. Give every signal an owner and a next action. Then record the outcome, including the interventions that failed.
A four-step account-health audit
- List the communication sources where client truth actually appears.
- Score satisfaction, activity, relationship coverage and growth separately.
- Attach one recent piece of evidence to every judgment.
- Turn each material change into an owner, deadline and review.
This is also the standard to apply to agentic software. A generated recap saves minutes. A cited intervention that catches an account drifting can save revenue. The second outcome is much harder because it depends on access, judgment, workflow and trust. Kaizan says its AI Helpers can draft outreach, prepare meetings, update systems and search for growth opportunities around the clock. The decisive detail is not that they act; it is whether the organization can see why they acted and decide where approval belongs.
That question becomes sharper when client calls and emails enter an AI system. Kaizan says it is SOC 2 Type II audited, offers EU and US data residency, keeps tenants separate and does not use customer conversations to train foundation models. Those are sensible foundations. Procurement teams still need to examine retention, permissions, model providers, redaction and the treatment of particularly sensitive accounts. Relationship intelligence concentrates information that clients may never have expected to appear in a score.
Adoption presents a quieter problem. The platform can connect to the inbox, calendar and meeting recorder, but useful context also lives in project choices, offline conversations and the reasons a team chose not to follow a recommendation. Kaizan says a standard onboarding takes two weeks and backfills 90 days of history. Buyers should use that period to test omissions, not simply admire populated dashboards. Put three disputed accounts through the model. Ask the people closest to each one what the score missed. The disagreements will reveal whether the system is learning the firm's reality or merely giving its data a polished surface.
Managers also have to decide how the information will be used. A relationship score designed to help a team can become a crude employee score if incentives are careless. Account directors may then optimize visible cadence, log performative activity or avoid difficult clients. The sensible unit of judgment is the account and its evidence, followed by a conversation about what support the team needs. Software can make an organization's blind spots visible; it can also make bad management more efficient.
A score cannot own the relationship
There is an appealing logic to a system that remembers every promise. There is also a danger in mistaking what is measurable for what matters. A delayed reply can signal dissatisfaction, parental leave, a crowded quarter or nothing at all. Sentiment is context, not verdict. The better design is one Kaizan's own evidence language points toward: software proposes, people inspect, and consequential action remains accountable.
The company is fifteen people, according to its April FAQ, and says it is hiring carefully. That makes its ambition look deliberately oversized. Calvert wants an organizational brain for every firm that serves clients. The near-term test is more concrete: can Kaizan become the place an account director checks before an important call, and can it catch a relationship change that a practiced human missed?
If it can, the product will have done something more useful than summarize another meeting. It will have made the soft infrastructure of client service visible without pretending the relationship itself belongs to the machine.
What is Kaizan?
Kaizan is an AI client-intelligence platform that combines calls, email, chat, documents and account data to score relationship health and help client-service teams act on risks and opportunities.
Who founded Kaizan?
Glen Calvert and Pravin Paratey founded the London company in 2022.
Does Kaizan replace a CRM?
No. Kaizan says it works alongside Salesforce, HubSpot and Pipedrive, then writes structured outputs back into existing systems.
What does CARE measure?
CARE separates client health into satisfaction or sentiment, account activity, relationship coverage and expansion or engagement growth.
How is it different from Pendo, Gainsight and Whatfix?
The tools overlap, but Kaizan starts with professional-services communications and relationship signals. Pendo emphasizes product behavior, Whatfix digital adoption and Gainsight broad customer-success operations.