The mandate is explicit: move Axi from agent-managed to AI-enabled servicing, run a contact-deflection roadmap prioritised by volume and cost, and evolve support into a retention function, all on DevRev. Below: the cost-to-serve read, how peer brokers service clients, the deflection thesis, each mandate mapped to a plan, and a working prototype of the console I'd run it from.
A broker's servicing cost scales with contacts, and most contacts are the same dozen questions: where is my deposit, why was my withdrawal delayed, how do I verify, what is margin, why did my MT4 disconnect. The job is to make the repeatable ones self-resolve, route the rest by client value, and use every contact as a retention signal instead of a cost. I'd run a deflection roadmap ranked by volume times cost, stand up tiered servicing tied to client segmentation, and wire complaints and churn signals back into product, all measured on resolution and retained clients rather than tickets closed. This page maps each of the four JD mandates to a plan, and links a working deflection console I built for it.
Axi (AxiCorp Financial Services, founded 2007, Sydney) is a global margin FX and CFD broker: 220+ instruments across FX, indices, commodities, shares and crypto, MT4 and MT5, plus the Axi Select funded-trader programme launched in 2023. It runs regulated entities under ASIC, the FCA, CySEC and the DFSA, so servicing is a regulated activity: complaints handling, appropriateness, and client-money rules all sit inside the support surface. Support today is human-led across live chat, email, phone, WhatsApp and a help centre, 24/5 in 13 languages. The role exists to keep that service quality while taking the cost and latency out of it with AI, and to make servicing pull its weight on retention. This is a servicing-transformation seat, not a helpdesk seat.
Retail-broker contact volume is dominated by a short list of repeatable, low-judgement questions. Each one handled by a human is a fixed cost with no retention upside. The chart below is my estimated contact-driver mix for a broker of Axi's shape (illustrative, from public support taxonomies), and the deflectability of each.
Mix is illustrative, modelled from public broker help-centre taxonomies (Axi, peers) and general fintech-support benchmarks. Real prioritisation would run off DevRev tags in week one.
The competitive read for this role is a servicing read, not a spreads read. Where do peers sit on self-service, AI, and proactive retention, and where is the gap Axi can take?
| Broker | Servicing posture | Gap Axi can exploit |
|---|---|---|
| IG scale leader | Deep help centre, 24/5 multi-channel, established chatbot triage; large mature CX org | Incumbent inertia. A leaner AI-native rebuild can beat it on resolution speed without the legacy stack. |
| CMC Markets | Strong self-service knowledge base, professional-grade tooling | Support skews power-user; less proactive lifecycle outreach to newer or mid-value clients. |
| Pepperstone | Fast human live chat is a core brand promise; award-winning support | Human-speed is the brand. AI-enabled resolution lets Axi match the speed at a fraction of the marginal cost. |
| IC Markets | 24/7 support, high-volume execution-focused base | Servicing is reactive and ticket-shaped; thin retention motion to build against. |
| eToro | Social/copy-trading community deflects some load; large retail base | Community is not account servicing; deposit, KYC and platform contacts still queue. |
| Axi the opening | 24/5 in 13 languages, multi-channel, moving to DevRev + agentic AI now | Mid-size and moving. Small enough to re-platform servicing fast, funded enough to do it on DevRev while peers sit on legacy tooling. |
Every large peer is bolting AI onto a decade-old support stack. Axi is doing the platform move (DevRev) and the AI move at the same time, at a size where you can actually re-cut the whole servicing model in quarters, not years. The prize is being the broker whose servicing is fastest and cheapest to run, and that turns the service desk into a retention channel while peers still treat it as a queue.
The JD names DevRev and "agentic AI." That is a specific bet: not a scripted FAQ bot, but agents that read the account, take a controlled action, verify the outcome, and escalate cleanly when they should not act. The maturity ladder I'd move Axi up:
DevRev's own benchmark for agentic deflection sits in the 40 to 60% range, higher as content gaps close. On Axi's contact mix, the deflectable top drivers (deposits, KYC, platform, admin) are ~70% of volume, so a disciplined roadmap can realistically clear the upper half of that band.
I would not chase the number blindly. A deflection that leaves a trader angry costs more than the contact. The target is resolved-and-satisfied, tracked as deflection rate × post-contact CSAT, with a guardrail on escalation-after-bot.
The JD asks to "evolve servicing into a retention-focused function with proactive outreach." In a CFD business, churn is expensive and largely predictable: first-deposit friction, a bruising early loss, a withdrawal that felt slow, silence after a funding issue. Servicing sees every one of those signals first. The move is to act on them.
| Churn signal (seen in servicing) | Reactive today | Proactive retention play |
|---|---|---|
| Deposit attempt fails or stalls | Client contacts us, maybe | AI detects the failed attempt, reaches out with the fix before they give up. |
| First withdrawal request | Processed silently | Trigger a reassurance + ETA touch; withdrawals are the top trust moment. |
| KYC stuck > 24h | Client chases us | Auto-nudge with the exact missing document; unblock funding faster. |
| Dormant after early loss | Nothing | Segmented education / risk-tools outreach, not a deposit push. |
| Repeat complaint or negative CSAT | Closed as a ticket | Route to a named human for save-the-relationship follow-up; log to product. |
The AI capacity freed from the standard tier is exactly what funds real human attention on the VIP tier. Deflection and high-touch are the same budget, re-allocated.
The posting groups the role into four mandates. Each one, turned into a plan, with where it is detailed on this page.
| JD mandate | My plan | Detailed in |
|---|---|---|
| 1 · Servicing leadership & operational excellence (global teams, service targets, DevRev data quality, compliance) | Own the SLAs for live chat, email and CSAT; make DevRev the single source of truth with clean tagging; run servicing inside the regulatory controls, not around them. | §06 |
| 2 · AI & automation-led transformation (agentic AI, deflection roadmap, agent-managed → AI-enabled, tiered servicing) | Rank contact drivers by volume × cost, deflect the top decile first, move up the agentic-resolution ladder, and tier servicing to client segments. | §04, §07, §08 |
| 3 · Client retention & commercial impact (proactive outreach, lifecycle, complaints, coaching leaders) | Convert churn signals seen in servicing into proactive plays; route complaints to product; lead and coach a multi-country leadership team on the new model. | ★, §09 |
| 4 · Strategic leadership (data-led exec reporting, transformation, org design, budgeting) | Run a tight servicing scorecard for the exec team, design the org around the AI-enabled model, and plan headcount and budget to the new cost curve. | §06, §09 |
How the servicing function runs day to day: the metrics I'd manage to, and the seams I'd own between AI, agents and the rest of the business.
Deflection rate, true-resolution rate, post-contact CSAT, first-response and full-resolution time, cost-to-serve per active client, escalation-after-bot, retention on serviced cohorts. Weekly to the team, monthly to the exec.
AI ↔ human handoff, servicing ↔ product (VoC), servicing ↔ compliance (complaints, appropriateness), servicing ↔ payments and KYC (the systems the agentic AI reads and acts on).
Multi-country leaders coached on the new model; agents re-skilled from queue-clearing to judgement work (disputes, VIP, retention saves); workforce plan tracks the falling cost curve honestly.
None of the reporting or AI works on dirty data. First priority inside the platform is a clean, enforced taxonomy: consistent contact reasons, disposition codes, and resolution states, so deflection is measured honestly and the AI trains on real signal. This is unglamorous and it is the thing that makes everything else true.
Prioritised by volume × cost, as the JD asks. Each wave ships a measurable slice of deflection, and nothing goes live without a CSAT guardrail and a clean escalation path.
The top ~43% of volume and the highest-anxiety moments. Give the AI controlled read access to payments and KYC state; answer "where is it / how long" instantly and accurately. Biggest, fastest win.
MT4/MT5 login and reconnect, password resets, detail updates, document re-submission. Agentic flows the bot can safely own end to end, with verification.
Margin, leverage, swaps, spreads, contract specs. Retrieval-grounded answers from Axi's own help content, in the client's language, deflecting the "how does X work" band.
Complaints, money-movement disputes, appropriateness and vulnerability signals, at-risk VIP accounts. The bot's job here is fast, full-context escalation, not an answer.
Flip from inbound to outbound: detect the failed deposit, the stuck KYC, the first withdrawal, and reach out before the contact happens (see ★).
Confidence threshold below which it escalates; post-contact CSAT and escalation-after-bot tracked per intent; a kill-switch per flow; human review of a sample every week. Deflection is earned, not forced.
Rather than describe the model, I built a working slice of it. The Axi Servicing Console is a browser prototype that shows the exact motion this role owns: a trader asks a real question, an intent engine classifies it, decides deflect / guide / escalate against a confidence threshold, answers from an Axi-grounded knowledge base, and a live dashboard tracks deflection rate, tier mix, cost saved and CSAT as you go.
It is a demo, mine, built for this application, not Axi code. The intent engine is a transparent retrieval + rules classifier so it runs with no keys and no data leaving the browser; in production this is where DevRev's Turing agents or an LLM slot in. The knowledge base uses public Axi facts. The metrics are simulated to show the reporting shape I'd manage to.
Facts are drawn from public sources as of mid-2026: the Axi job posting and website, Axi's help centre and regulatory disclosures (ASIC, FCA, CySEC, DFSA), DevRev's public material on agentic support and ticket deflection, and public reviews and comparisons of peer CFD/FX brokers (IG, CMC, Pepperstone, IC Markets, eToro). The contact-driver mix and all metrics on this page and in the demo are illustrative, modelled from public support taxonomies and fintech benchmarks, not Axi's internal figures. This is an unsolicited blueprint prepared as interview homework; happy to walk through any section.
Axi: axi.com/int · help.axi.com
DevRev deflection: devrev.ai, ticket deflection
Role: Head of Client Success (Greenhouse posting, 2026)
Peer servicing: public broker reviews & help centres (2026)
Independent blueprint for the Axi Head of Client Success role · 2026 · edwardtay.com