Custify’s Post

Customer engagements start months before an actual meeting. Some of the most important work happens before anyone even walks into a room. In the months leading up to next week's Google Cloud Global Customer Advisory Board (gCAB), our Strategic Customer Engagement (SCE) team spends time understanding what board members are navigating in their businesses, where they are feeling pressure and which conversations would be most valuable when we bring the group together. This work shapes everything that follows: the agenda, the questions we put on the table and the Google leaders and experts we bring into the conversations. It’s one of the things I value most about this work. The goal isn’t to fill an agenda. It’s to make sure the time we have together is focused on the issues customers have told us matter most.

Talent alone doesn't compound. The system does. A co-sell motion either builds on itself or resets every quarter, and the difference is almost never the people. It's whether the operating principles were in place before the work started. After 30+ years running alliance and co-sell programs across AWS, GCP, Snowflake, and Databricks, both Shawn James, COO, and Steve Gordon, Managing Director, have watched motions fail the same three ways: assumption over foundation, diffused ownership, and tools chosen before the workflow was proven. So before we at AlignBiz enter any cloud co-sell engagement, three rules hold. Even when the client wants to move faster. Even when holding them slows the start or costs a deal. 1. We audit the existing hyperscaler relationships first. No motion design before we understand what's actually in place inside AWS, GCP, Snowflake, or Databricks. You can't build on a foundation you haven't looked at. 2. We name a single accountable owner before any partner is engaged. Shared ownership means the motion stalls. Hyperscaler relationships need daily attention from one person who lives the motion, not a committee that reviews it. 3. We don't let platform selection precede process definition. Technology encodes a working motion. It doesn't create one. If the tool is chosen before the workflow is proven, you're automating a guess. These aren't preferences. They're what separates a motion from a mechanism. If your co-sell motion keeps resetting, the operating principles are probably missing. Connect with AlignBiz to audit what's actually in place before the next engagement begins.

Proud to share that Custom Agents, one of the products I’ve been designing at Zendesk, is featured in a case study from Anthropic. Every business has work that’s specific to them: return policy exceptions, warranty claims, billing edge cases. Off-the-shelf AI doesn’t know any of it, so handing that work to AI has been hard to trust. Custom Agents is our answer: a way for anyone to build AI agents that follow their exact playbook. A few numbers from the first seven weeks of early access: → 1M+ agent executions → Up to 10% increase in automated resolution → Up to 80% reduction in human handle time The part I’m proudest of is helping people feel confident handing real work to agents they buid themselves. Huge thanks to the team who made this happen, and to the customers who built with us from day one: Abhiney Kathuria, Brett Schuenemann, John Ferlazzo, Jeremy Kirkwood, Harrison Shaw, Setu Chokshi, Rahul Trikha, Harshit Sethi, Rikish Kamboj, Dana Ma, Liz Robins and many more. Read the full story: https://lnkd.in/guzQ4WZs #ProductDesign #AI #CustomerService #Zendesk

"Agentifying" your business with custom agents for your specific challenges, whilst maintaining and adhering to company specific guardrails, is not easy. Many customers ask Zendesk for an agentic framework that balances self-serve flexibility with traditional enterprise-level B2B software requirements. Zendesk runs its infrastructure on Amazon Web Services (AWS), so it serves Claude through Amazon Bedrock. Zendesk keeps model processing in the AWS hosting regions it already uses, which helps it meet customers’ data-residency and governance requirements. If you'd like to better understand use cases such as: a) how an non-engineer, information worker can use Natural Language (literally speak the construction of the agent) to create an agent that speaks to third party applications or b) customer/employee service triage (intelligent routing and completion of multiple and related tickets) - then here is a story of how Zendesk with #AWS delivered custom agents for Customer and Employee Service on Claude https://lnkd.in/ecYU43MW cc: Luke Donnelly | Lucinda McCaffrey | Sarah Al-Hussaini | Max Cunliffe | Peter Tailby | John Kelleher | Lee Fairhurst | Eoin O Cuilleanain | Owen McConnell | Tanja Hilpert | Björn Bauer | Patrick Renger | Christopher Baxter | Dmitry Osoblivets | Stephan Mey | Michael Chaouat | Thomas Verschoren | Thor Newman Andersen | Alberto M. Becerra | karine GARCINI | Carlos Llorca Rodriguez del Valle | Jasita Lalloo (B.A., B.A.I., MBA)

Zendesk x Claude x AWS = 💥 If you’re not using custom agents on Zendesk already, it’s time to get started!

Your users aren’t ignoring your docs. They just can’t find them. Every SaaS support team knows this ticket: “How do I export my data?” The answer has been in the help center for months. So the team pastes the same links all day, and everyone wonders why ticket volume keeps climbing. Why it happens: - Articles are written in your words, not your customers’ words - Search doesn’t match how people actually phrase the problem - The help link isn’t there when the user gets stuck How to fix it for good: 1.Pull the last 30 days of tickets and tag every “already documented” question. 2.Rename your top articles using the exact phrases customers use in tickets. 3.Put the answer where the question happens: in-app tooltips, onboarding emails, and suggested articles on the contact form before someone submits. 4. Build a macro for each repeat question and track which articles get sent most. 5. Review it monthly. If the same article keeps getting sent, the page or the product needs fixing. 6.Having docs isn’t the same as having docs that work. Which question does your team answer the most even though it’s already in your help center? #SaaS #CustomerSupport #CustomerExperience #HelpCenter #SupportOps #Operations

Big news for Abstrakt. We’ve been acquired by TCN. TCN provides AI-powered cloud contact center software focused on automation, compliance, and customer communications. For us, the fit is pretty straightforward. We help agents perform better during the conversation and give managers better visibility into what happened afterward. TCN gives us the opportunity to bring that technology to more contact center teams as part of a larger platform. We’re excited about what’s next! Read the full announcement: https://lnkd.in/ez-haVgF

This is a big deal for anyone building agents that touch real customers, not just internal teams. For years, the security model for "who can use which tool" assumed one company, one set of employees, one admin console. Agentic AI breaks that assumption fast: the moment you're serving multiple external tenants from a single agent deployment, you need per-tenant isolation — not just per-user permissions. What stands out to me in Andres and the team's work here is the framing: every tool call bound to end user, tenant, andagent, with kill switches scoped down to the tenant or integration level. That's the difference between "we secured our agent" and "we secured our agent for a multi-tenant SaaS world" — and it's exactly the kind of problem identity has always solved, just applied to a new layer of the stack. If you're building customer-facing agents and haven't thought through what happens when one tenant's credentials are compromised, this is worth a read. 🔗 https://lnkd.in/dQXekzSe This is shaping up to be a core piece of #Auth0's Agentic Runtime.

A concern I keep hearing from practitioners evaluating Datadog: they want deeper visibility into their apps, services, and infrastructure, but worry about the bill as cardinality and telemetry volumes grow It’s a fair concern. I also think the conversation needs to go deeper into how observability data is collected, retained, and used. That applies regardless of whether you use Datadog Would recommend this write-up from Zendesk’s engineers. It goes into the weeds of how they controlled costs while preserving the visibility needed for troubleshooting TL;DR: 1. Make costs visible by service and team so you know where to focus 2. Remove redundant logs at the source. Cleaning up duplicates reduced log volume by 75% for one high-volume service 3. Adjust trace sampling based on service volume and criticality, while preserving coverage of critical paths 4. Review ingestion and indexing separately. Excluding logs from indexing still leaves you paying to ingest them Zendesk flattened cost trends across APM and logs while keeping the operational data their engineers needed https://lnkd.in/g-Cz-xnq

What makes a technology ecosystem operational, not just connected? (2 mins read) This week’s ecosystem signals center on commercial measurement, workflow discovery, architecture and partner readiness. 1. AWS turns partner performance into program evidence Amazon Web Services (AWS) makes attributed revenue a direct measure of partner performance. Partner Revenue Measurement connects that revenue to partner products and services, with funding qualification tied to attributed revenue starting in 2027. Commercial contribution is becoming measurable enough to shape how the ecosystem allocates investment. https://lnkd.in/gQ6jXz2a 2. TwelveLabs turns agent access into workflow discovery TwelveLabs exposes Jockey to Claude and ChatGPT as a callable video agent. Agent access surfaces new workflow patterns around video, pushing Jockey beyond search and summary. When discovered workflows extend capabilities, usage shapes what the product becomes next. https://lnkd.in/gTyhzRap https://lnkd.in/g9QbnNif 3. Imagine makes multi-vendor architecture an operating model Imagine Communications positions DMF and MXL around a broadcast environment where specialized systems can participate without forcing the workflow into a single-vendor stack. That shifts interoperability from a connector problem to an architectural decision. The value is in creating stable exchange and control layers that let each system do the job it is best at. https://lnkd.in/gaPePAtU 4. AI readiness reframes the MSP market Dina Moskowitz reframes the MSP market by making AI readiness a key component of partner profiles. PartnerOptimizer, Inc. finds that, across a 46,000-MSP cohort, more than 8,700 are AI-forward, while MSP 501 winners show a much higher concentration of named AI practices and products. Partner recruitment turns from broad category matching toward identifying firms already demonstrating observable readiness and adoption capacity. https://lnkd.in/gbegXuvv Where the value is moving: ➡️ Partner performance becomes funding evidence. ➡️ Agent access becomes workflow discovery. ➡️ Specialization without stack lock-in. ➡️ Readiness directs partner investment. The value is moving into the mechanisms that decide what gets used, what gets funded and what gets built next.