Full-Time

Senior Platform Architect

Updated on 9/5/2026

Zendesk

Zendesk

5,001-10,000 employees

Subscription CRM software for customer service

No salary listed

London, UK

Hybrid

Hybrid work requires working remotely for part of the week and meeting in person at a Zendesk office.

Category
Solution Engineering (1)
Required Skills
LLM
SharePoint
Microsoft Azure
JavaScript
React.js
Node.js
ServiceNow
RAG
AWS
Vue.js
REST APIs
Confluence
Data Modeling
HTML/CSS
OAuth
Google Cloud Platform

Get referred to Zendesk

See people who can refer or advise you

Requirements
  • At least 8 years of experience in enterprise system or solution architecture, technical presales, or professional IT consulting services.
  • Exposure to or hands-on experience architecting and deploying generative artificial intelligence solutions, prompt engineering, Retrieval-Augmented Generation frameworks, and agentic artificial intelligence reasoning systems.
  • Advanced expertise with RESTful APIs, JSON implementations, web services, database design, data modeling, and Model Context Protocol configurations.
  • Deep familiarity with AWS, Azure, or Google Cloud Platform environments and experience orchestrating enterprise-wide data using integration-platform-as-a-service applications.
  • Comprehensive knowledge of Single Sign-On configurations, OAuth, advanced data privacy frameworks, and token-based artificial intelligence authentication capabilities.
  • Solid understanding of HTML, CSS, JavaScript, and modern SaaS application development architectures.
  • Ability to influence senior stakeholders, translate technical artificial intelligence concepts into business outcomes, and deliver technical presentations both onsite and remotely.
  • Ability to prioritize multiple complex, concurrent transformation accounts.
Responsibilities
  • Partner with sales engineers, account executives, and business consultants to drive strategic discussions and communicate technical architecture, solutioning, and the business value of artificial intelligence to C-suite and director-level executives.
  • Drive annual recurring revenue through automated resolutions by conducting strategic technical discovery, scoping, and managing complex proofs of concept for expansion and cross-sell opportunities.
  • Demonstrate Zendesk headless technology and help customers embed Zendesk into workflows inside and outside the core platform using Model Context Protocol and agent-to-agent technology.
  • Master and deploy Zendesk Agent Builder and Action Builder to architect multi-step, agentic workflows targeting at least 80% automated resolution.
  • Architect modern integrations connecting the Zendesk platform to external artificial intelligence platforms and enterprise applications using the Zendesk Model Context Protocol Server, APIs, and integration-platform-as-a-service solutions.
  • Apply scalable enterprise integration patterns.
  • Use Zendesk's Context Graph and Knowledge Graph connectors to integrate unstructured, federated enterprise data sources such as SharePoint, Confluence, and Google Drive into customers' artificial intelligence ecosystems without data migration.
  • Lead the technical handoff between presales and postsales implementation teams to support long-term artificial intelligence deployment success, continuous quality assurance through artificial intelligence quality scores, and ongoing expansion.
  • Serve as the organizational subject matter expert on Zendesk platform security, planetary scalability, custom data models, and Zendesk Copilot implementations.
  • Partner with Product Engineering and Product Marketing as a feedback loop to shape the future roadmap of Zendesk artificial intelligence architecture.
Desired Qualifications
  • Previous experience managing large-scale implementations of enterprise-class customer service software or ITSM platforms.
  • Experience using advanced diagnostic analytics systems for root-cause and trend analysis.
  • Familiarity with modern JavaScript frameworks such as React, Vue, or Node.js for custom application development.
  • Practical experience using artificial-intelligence-augmented tools such as Claude Code or Cursor.
  • Strong sales acumen and prior experience in technical presales or solutions architecture within an enterprise SaaS or customer-experience environment.
  • Experience with Salesforce Service Cloud, ServiceNow, Intercom Fin, or Freshworks.

Zendesk provides cloud-based customer service and support software that helps businesses manage customer interactions. It offers a suite of tools (ticketing, live chat, analytics) as part of a subscription-based CRM platform designed to streamline support workflows across industries. The product works as a scalable cloud solution: organizations choose pricing tiers, implement ticketing and chat in their support channels, track performance with analytics, and use training and certification resources to maximize use. What sets Zendesk apart is its broad, adaptable platform tailored for companies from startups to enterprises, extensive integrations, and a clear focus on service-first CRM, with a recurring revenue model built on subscriptions. The company’s goal is to help organizations improve customer service operations, boost efficiency, and enhance customer satisfaction and loyalty by making support processes easier to manage.

Company Size

5,001-10,000

Company Stage

IPO

Headquarters

San Francisco, California

Founded

2007

Get referred to Zendesk

See people who can refer or advise you

Simplify Jobs

Simplify's Take

What believers are saying

  • May 19, 2026 autonomous-service launch expands AI pricing and widens enterprise wallet share.
  • September 3, 2026 rollout gives all customers advanced AI agents, accelerating adoption.
  • June 4, 2026 Forethought availability and 40+ connectors strengthen cross-system resolution.

What critics are saying

  • Zendesk will shut off legacy AI agents and bot builder on December 10, 2026.
  • March 2026 layoffs and repeated cuts signal ongoing restructuring under PE owners.
  • Salesforce, ServiceNow, and niche AI startups attack Zendesk’s core support workflow moat.

What makes Zendesk unique

  • Zendesk’s March 11, 2026 Forethought deal deepens autonomous service workflows across channels.
  • Its May 19, 2026 Agent Builder ships no-code custom AI agents with governance.
  • The platform combines service CRM, voice, employee service, and MCP-connected workflows.

Help us improve and share your feedback! Did you find this helpful?

Benefits

Hybrid Work Options

Flexible Work Hours

Growth & Insights and Company News

Headcount

6 month growth

2%

1 year growth

1%

2 year growth

0%
Bear Systems
Sep 3rd, 2026
NetSuite-Zendesk sync: eliminate credit memo reconciliation errors.

NetSuite-Zendesk sync: eliminate credit memo reconciliation errors. Manual credit memo reconciliation between NetSuite and Zendesk costs enterprises 15+ hours weekly. Automating this sync reduces errors by 95% and cuts audit time from days to minutes. Every month, finance teams at mid-market enterprises waste 15-20 hours reconciling credit memos between NetSuite and Zendesk. The process begins with exporting CSV files from both systems, cross-referencing ticket IDs against refund records, and manually adjusting GL entries when discrepancies arise. A single misaligned credit memo - whether due to a typo in NetSuite's refund application or a Zendesk agent's incorrect ticket closure - can trigger a 3-day audit cycle to trace the root cause. For companies processing 500+ refunds monthly, this isn't just tedious; it's a systemic bottleneck that delays month-end close by an average of 2.3 days, according to NetMint's connector documentation. The breakdown isn't just in the volume. It's in the handoffs: NetSuite's refund application module doesn't natively track Zendesk ticket statuses, while Zendesk's credit memo fields lack NetSuite's GL impact granularity. When a customer disputes a refund in Zendesk, agents often create duplicate tickets or mislabel credit memos, forcing finance to reconcile the same transaction twice. The result? A spreadsheet maze where 1 in 5 credit memos requires manual intervention, per SyncApps' integration case studies. For a company with $50M in annual revenue, the labor cost of manual reconciliation alone exceeds $18,000 annually - assuming $35/hour for a mid-level accountant and 15 hours weekly. But the real expense is opportunity cost. Each hour spent firefighting credit memo discrepancies delays financial reporting, eroding investor confidence and complicating tax filings. In industries like SaaS or e-commerce, where refund velocity correlates with churn risk, delayed reconciliations can mask cash flow leaks until they're severe enough to trigger an audit. The secondary costs are harder to quantify but no less damaging. Spreadsheet errors propagate into NetSuite's revenue recognition reports, risking SOX compliance violations. Meanwhile, Zendesk agents waste 8-10 minutes per ticket resolving refund disputes that could've been auto-closed with accurate credit memo data. Multiply this by 1,200 tickets monthly, and the cognitive overhead alone justifies automation. As Zapier's NetSuite-Zendesk integration guide notes, even 'minor' discrepancies can cascade into compliance risks when left unchecked. Bear Systems' integration leverages NetSuite's SuiteTalk REST API and Zendesk's Support API to create a bidirectional sync that propagates credit memo statuses in real time. The workflow begins when a refund is processed in NetSuite: the system pushes the credit memo's ticket ID, amount, and GL impact to a middleware queue (AWS SQS), where Bear's AI agent validates the data against Zendesk's ticket status. If the ticket is closed with a 'refund issued' tag, the agent triggers a confirmation in NetSuite's refund application module, eliminating the need for manual GL adjustments. For Zendesk-side updates, the process reverses: when an agent marks a ticket as 'credit memo issued,' the AI agent cross-references the ticket ID with NetSuite's refund records. If no matching credit memo exists, the agent flags the discrepancy for finance review via Slack/Teams. This eliminates the 'orphaned refund' problem, where credit memos sit unapplied in NetSuite for weeks. The integration also enforces data schema alignment - NetSuite's refund fields map directly to Zendesk's custom credit memo attributes, ensuring consistency without manual mapping. For a $75M revenue company processing 800 refunds monthly, Bear's integration delivers $42,000 in annual labor savings - assuming $40/hour for finance staff and 20 hours of weekly reconciliation time. The error rate drops from 20% to under 1%, cutting audit cycles from 3 days to under 30 minutes. More critically, the integration reduces month-end close delays by 2.1 days on average, enabling faster financial reporting and board presentations. The strategic value extends beyond cost savings. With credit memo data flowing seamlessly between systems, finance teams can now generate real-time refund analytics - e.g., tracking refund velocity by product line or agent - to identify churn risks proactively. Compliance teams benefit from automated audit trails, as every credit memo sync is logged with timestamps and user attribution. Compared to generic connectors like Zapier's NetSuite-Zendesk integration - which lacks AI-driven discrepancy resolution - Bear's solution reduces false positives in reconciliation by 70%, per internal benchmarks. A full rollout takes 6-8 weeks, starting with a 2-week discovery phase to map NetSuite's refund workflows and Zendesk's custom fields. Prerequisites include NetSuite's Advanced Financials module (for GL-level credit memo tracking) and Zendesk's Professional tier (to enable custom credit memo attributes). The integration requires API rate limit adjustments in both systems - NetSuite's SuiteTalk API caps at 5,000 calls/hour, while Zendesk's Support API allows 100 calls/minute - so throttling logic is baked into the middleware. Common pitfalls include unstandardized ticket IDs (e.g., NetSuite's refund ID vs. Zendesk's ticket number) and misaligned refund amounts due to currency conversions. Bear Systems mitigates this with a pre-sync validation step that flags mismatches before they propagate. Another risk is agent pushback: Zendesk teams may resist adopting new credit memo workflows if the UI changes aren't intuitive. To counter this, Bear provides a pre-built Zendesk macro that auto-populates credit memo fields based on NetSuite's refund data, reducing agent friction by 60%. Generic connectors like SyncApps or Zapier can sync NetSuite and Zendesk, but they lack the AI-driven discrepancy resolution and compliance logging required for credit memo reconciliation. NetMint's connector, for example, handles basic data sync but doesn't enforce GL-level accuracy or provide real-time audit trails. For enterprises where refund errors can trigger compliance audits or investor scrutiny, a purpose-built integration isn't just faster - it's lower-risk. Bear Systems' approach combines NetSuite's financial rigor with Zendesk's operational agility, ensuring that credit memo data is both accurate and actionable. If your team is still wrestling with spreadsheets or firefighting refund disputes, the next step isn't another tool - it's a conversation about how to eliminate the manual loop entirely. Audit your current credit memo reconciliation process this quarter; if it's costing more than $15,000 annually in labor or delaying financial close, it's time to talk. Free Business Automation Audit Discover how much time and money your team can save by automating manual workflows and integrating enterprise systems.

Palo Alto Networks
Sep 2nd, 2026
Building securely from day one: Palo Alto Networks partners with the Zendesk Startup Program.

Building securely from day one: Palo Alto Networks partners with the Zendesk Startup Program. Sep 02, 2026 A secure workspace for your team - so you can focus on building, not on what could go wrong. Every customer support agent starts their day the same way: opening a browser, logging into Zendesk, and handling sensitive customer interactions. For most early-stage teams, that workspace is completely unprotected. The browser is where your people work, where your customer data lives, and where most security incidents begin - yet it's the layer that gets secured last, if at all. AI is now part of everyday work. Your team is using agents and AI-powered workflows to move faster. But as your AI usage scales, so does your exposure: sensitive customer data, keys, source code, credentials, and proprietary information can flow into unvetted AI tools without anyone noticing - including you. And the browser is where all of it happens. The startup ecosystem thrives when founders can focus on building. That's Palo Alto Networks partners with the Zendesk Startup Program to provide early-stage teams access to the same browser security used by large enterprises - so startups are not forced to choose between preserving runway and protecting their customers' data. A secure workspace, not just a secure browser. Picture this: a support agent on your team gets a convincing phishing email that looks like a familiar notification. They click the link, land on a spoofed login page, and enter their credentials. Within hours, an attacker has access to your support system, and every customer conversation, ticket, and piece of sensitive data inside it. You find out when a customer calls to ask why your support team reached out, asking for payment details. This isn't a hypothetical. It's how many small business breaches actually play out - and it's exactly what PBB stops. Palo Alto Networks, Inc. use Palo Alto Networks's Precision AI-powered security to recognize the destination as malicious and block it before the page ever loads. No credentials entered. The agent sees a warning and moves on with their day. That's the core idea behind Prisma Browser for Business: a secure workspace that sits at the intersection of work, AI use and threats - the browser. PBB wraps your entire browser session in enterprise-grade security, protecting every app, every tab, and every interaction, without requiring your team to change how they work. For Zendesk users specifically, PBB can be configured with Zendesk as a dedicated workspace app - automatically applying the right security policies to everything Zendesk-related the moment your team opens it. Phishing attempts are blocked. Data leakage from support conversations is prevented. And your agents work inside a protected environment without any extra steps on their end. The result: your team gets a seamless experience, and you get end-to-end visibility and control over how Zendesk is accessed across every device on your team. The security gap startups can't afford to ignore. When you're building a company, security can feel like something to tackle later - after the product ships, after the team grows, after the next funding round. But phishing, malware, and data leakage are not enterprise problems. They're startup problems too - and they can be significantly more damaging when you don't have a dedicated security team to catch them. Three risks in particular are growing fast for early-stage teams: * AI tools without guardrails. Your team is almost certainly using AI tools - Claude, ChatGPT, Copilot, and others - to move faster. But without controls in place, sensitive customer data, internal documents, and proprietary information can flow into these tools without anyone noticing. PBB lets you set guardrails around how AI tools are used in the browser, so your team can stay productive without inadvertently exposing data they shouldn't. * Unvetted browser extensions. Extensions are one of the most overlooked attack surfaces in any organization. A single malicious or compromised extension can silently capture credentials, intercept data, or give an attacker persistent access to your environment. PBB gives you control over which extensions can be installed across your team's browsers, and alerts you when a suspicious or malicious extension is detected - before it can do damage. * Phishing and malware. The browser is the primary delivery mechanism for both. PBB blocks malicious sites and stops malware downloads in real time, protecting your team and your customer data at the point of risk. PBB brings all of this together in a way that is simple to deploy and easy to manage - no security expertise required, no complex infrastructure to stand up. Built for where startups actually are. The Zendesk Startup Program is designed to support founders from their earliest stages through scale - including when security becomes a core business priority. Startups eligible for the program can access PBB as a practical way to evaluate enterprise-grade workspace security without a large upfront commitment. It's a way to build securely from the start, preserve runway, and scale your operations with confidence. As a rapidly growing startup, you shouldn't have to choose between moving fast and staying secure. * The browser is your biggest security gap - and it's where startups get hit first. Phishing, credential theft, malicious extensions: these aren't enterprise problems. They're startup problems, and they land in the browser before you even have a security team to respond. * Security doesn't have to slow your team down. PBB applies enterprise-grade protection automatically - no behavior change required from agents, no complex infrastructure to stand up. Your team works exactly as they do today, just inside a protected environment. * Build securely from day one, without the day-one security budget. Leverage protection large enterprises rely on - so you don't have to choose between your runway and your customers' trust. Ready to get started? Startups participating in the Zendesk Startup Program are eligible to apply for PBB as part of their program benefits. Visit the Prisma Browser for Business dedicated landing page for Zendesk to take advantage of its partner's promotion.

Brij
Aug 31st, 2026
Brij x Zendesk integration.

Brij x Zendesk integration. Audrey Buck August 31, 2026 Integrations * For customers who never placed an order with the brand directly, their Brij registration becomes the only proof of purchase that exists. * The Brij + Zendesk integration displays a customer's product registration, warranty status, purchase details, and receipt upload directly in the Zendesk ticket sidebar, so agents can confirm ownership in the first reply. * The app matches automatically by comparing the ticket requester's email to a Brij registration profile, so nothing displays unless a real match exists. * Agents can verify retail and marketplace purchases directly inside Zendesk, closing warranty and ownership questions in one exchange. When a customer buys directly from your site, your support team can pull up the order in seconds - name, email, SKU, shipping status, all sitting in the same system as the ticket. When that same customer buys at Target, Walmart, or on Amazon, there's no order to look up. Your agent is staring at a blank field, and the only way to move the ticket forward is to ask the customer to dig up a receipt, screenshot it, and wait. That back-and-forth is the retail black box showing up in your help desk. The purchase happened, but your support stack has no record of it. Brij closes that gap by turning retail and marketplace purchases into registration data your team already trusts, and now that data lives where support agents actually work. The Brij + Zendesk integration puts it directly in the ticket sidebar. What is Zendesk? Zendesk is one of the most widely used customer service platforms that helps businesses manage support conversations across email, chat, phone, social media, and other channels. It gives support teams the tools to resolve issues faster, organize customer interactions, and deliver a more consistent service experience at scale. Most omnichannel brands already run their support desk through it. What they're usually missing is any visibility into the 80% of sales happening outside their own site. Verify your retail buyers directly in Zendesk. Brij's registration is often the only proof-of-purchase record that exists for non DTC customers: they scanned the code on the package, registered the product, and uploaded the receipt. That record now appears in the ticket, so an agent can confirm ownership in the first reply instead of asking for a receipt and waiting two days. * No more "can you send a photo of your receipt?" Agents see registration and purchase details the moment a ticket opens. * Warranty status at a glance. If a customer registered a product through Brij, the sidebar shows it, no separate lookup required. * Faster first-response time. Agents resolve retail-purchase questions in the same motion as DTC ones, instead of pausing the ticket to wait on the customer. * PII stays controlled. Access can be restricted by role or group, so only the agents who need registration data can see it. How the Brij + Zendesk integration works. * Connect your Zendesk account inside Brij's integration settings. * Choose what surfaces. Configure which customer, registration, product, and warranty registration details should appear in the sidebar. * Install the Brij Customers app as a private app in Zendesk Admin Center (not a marketplace listing), using the API key and base URL your Brij account manager provides. * Set access controls, optionally restricting the app to specific agent roles or groups. * Open a ticket. If the requester's email matches a Brij registration, their product, warranty, and receipt data appears in the sidebar automatically. If it doesn't match, nothing displays, no dead ends, no false positives. Use cases. * Warranty disputes. An agent confirms a customer registered their product and what warranty terms apply. * Order verification for retail purchases. A customer says they bought at Target; the agent confirms it instantly instead of taking their word for it. * Proactive education. Agents reference the setup guide or survey responses a customer already interacted with, instead of re-explaining from scratch. * Escalation triage. Support leads use registration and warranty status to decide whether a ticket needs a refund, a replacement, or just an answer. * Cross-team handoff. The same data Brij surfaces in Zendesk also feeds marketing and retention, so a support interaction doesn't dead-end as a one-off fix. It's part of the same customer success loop. Build a stronger support experience. Support agents can't help customers they can't see. The Brij + Zendesk integration gives your team the same proof-of-purchase confidence for a Target sale that they already have for a DTC order, right inside the ticket they're already working. Brij turns those retail purchases into identity your whole stack can use, support included. Explore the full list of Brij Integrations, or book a demo with Brij to see the Zendesk sidebar in action. Faq. What data does the app show in a ticket? Brij registration data matched to the ticket requester's email - product registrations, warranty information, purchase details, the receipt upload if one exists, and any survey answers your registration flow collects. If the requester has no Brij registration, or their Zendesk email differs from the one they registered with, the panel shows nothing. Is this on the Zendesk Marketplace? No. It installs as a private app. Your Brij account manager provides a .zip file and your Brij API Base URL, and a Zendesk admin uploads it under Apps and integrations | Zendesk Support apps | Upload private app. Can I limit which agents see customer data? Yes. Zendesk role and group restrictions apply to the app, so you can scope it to the teams handling warranty and product tickets and keep registration PII away from everyone else. Can I add a Zendesk chat widget to a Brij experience? Yes, though that's separate from this app. Paste your Zendesk chat snippet into Settings | Custom Code in Brij and the widget loads across your experiences, so a customer can start a conversation without leaving the flow.

Claret Capital Partners
Aug 18th, 2026
EdgeTier raises funding from Claret Capital Partners for agentic AI customer service platform

Claret Capital Partners has invested in EdgeTier, a Dublin-based AI-powered customer service intelligence platform, through its Fund IV. Founded in 2017, EdgeTier employs over 70 people and serves more than 50 enterprise clients across Europe and the US, including TUI, Abercrombie & Fitch, Ryanair and Wolt. EdgeTier's platform analyses customer conversational data in real time, providing insights into emerging issues and agent performance trends. The agentic AI-powered system integrates with major customer support platforms such as Salesforce, Zendesk, Genesys, NICE, Intercom and LivePerson. CEO Shane Lynn said the investment will help EdgeTier scale its platform and expand globally. EdgeTier's existing investors include Smedvig, ACT Ventures and Episode 1.

CXFocus Magazine
Aug 17th, 2026
Amplify your strengths not your sins - Sealy on the reality of AI in CX.

Amplify your strengths not your sins - Sealy on the reality of AI in CX. When Michael Chen joined premium Australian mattress manufacturer Sealy Posturepedic nearly 7 years ago, centralising the fragmented CX function was his first priority. Laying the right foundations with the CX team, before introducing process automation tech, was most important. "People, culture, processes, come first and then and only then the technology. If you have people problems, gaps in your existing teams and culture, AI will only help you to experience those things faster," says Michael Chen. Sealy's Head of Digital and CX. "But if you have helped your team to understand that the technology will help to make them more efficient, that AI is a tool to help them, not replace them, then your CX people will deliver measurably better customer service outcomes." Sealy's AI journey. Sealy's AI journey began during the Covid lockdowns. Incoming customer calls and messages "skyrocketed" and response times blew out. The Executive Team initially considered hiring more people but movement restrictions meant that was not doable. "So we landed on tech solutions, including co-pilot agents, that allowed us to do more with our existing team. From there, our vision broadened from just wanting to clear the backlog to creating a solution for our operating model that we can scale up and down as required and make us more efficient," Chen explained. "It wasn't all about what technology are we buying, more about how we can take our team with us on the journey because the CX model that had worked for us for the previous 20 years no longer did. We went looking for a strategic solution that was going work for us for the next 20 years." That led Sealy to adopting the Zendesk platform and its co-pilot agents before, very recently, integrating Zendesk's AI agents into their customer-facing operations. "Sealy's initial results with Zendesk's AI agents are very encouraging. Customer retention is up 100%. Our team capacity is up 30% and we maintained a 90% CSAT score while rolling out AI," according to Michael. Judgement and empathy. The judgement, empathy, the high values skills of the human agents, are central to this success, he believes. "It's human agents with the right mindset who have the ability to really connect with customers, solve problems, reassure them and build trust. Our people kept our CSAT at 90%. People who can really focus on people and understanding customer needs, they're going to become more important to us in future." The company is still relying heavily upon CSAT to track customer satisfaction. "Immediate data is less important because we have a lot of data points in the past, so we look at the overall trend data going through but its early days for us so we will see," Chen notes. Sealy have deliberately pursued AI implementation in small steps, making it easier for the Executive team to measure ROI and manage operational adaptation, including training. "Start small, don't do everything at once," is Chen's advice. "Focus on good process culture. Launch one or two things. Get your agents comfortable, get the business comfortable, get some runs on the board, this will help you do the rest of the implementation a lot better too," Michael says. Sealy are currently using AI agents for chat customer engagement. Voice enquiries are still going to a 100% human contact centre, less than a month into the implementation phase. Chen appreciated having a technology service partner who played an active role in Sealy's journey. "When we started deploying AI tools, we wanted to see how they could help us with our existing processes. You want to make sure the tools are helping the human agent. Getting wins at the beginning, gets the human buy-in you need to succeed." "Co pilots have many, many capabilities. So what we initially selected was what we thought would benefit the agents. When we launched co pilots, we did not have a clear expectation but with call times immediately reduced by 2-3 minutes this allowed us to redeploy those extra resources. We were able to show the Executive Team 'yeah we've spent this money on technology but we're going to get a return on it," Chen recalls. "With a co pilot, everything is shared with human agents but with AI the machines will respond to customers without a human watching over them. So, for the two weeks of AI implementation, we closely monitored what the AI agents were doing during the day and turned them off at night. We've only just moved to leaving the AI agents on over the weekend." "It's constant learning. Constantly adjusting, optimising, the technology while you are learning about it and the technology itself is improving so quickly," Chen says. "Initially, implementation is very resource intensive but we are already seeing 40% to 50% of chats are automated which is better than we initially anticipated. If people want to know 'do you have stock of an item' or how is their ticket progressing, our AI chatbots can give customers that information right away at 3am. Or they can get that information from our website, whichever they prefer." Adopting a hybrid-AI system has also highlighted the importance of having accurate, detailed and up to date information on the website and other company digital information channels that AI agents rely upon to generate answers for customers.