FDE Sprint
ONE PROCESS
One process, from baseline to live.
- Length
- 6 weeks
- People
- 1 FDE
- Scope
- One process, baseline to go-live
| Comparison | Generic AI consultants / trainers / SaaS vendors | EKel AI Advisory |
|---|---|---|
| Pricing model | Hourly / one-off / annual subscription | Fixed-price audit + monthly retainer |
| Time to first result | 2–3 months after proposal | Audit report delivered in 2 weeks |
| Tied to operating KPI | Consulting report is the deliverable | Every recommendation tied to a quantified metric |
| Ability to ship | Advice only, you find a vendor | Same team continues into an Agentic Coding sprint |
| Exit mechanism | Annual contracts with cancellation fees | Audit refundable / retainer 30-day notice |
| After delivery | They leave | Fractional AI Lead keeps it running and expands it |
// NEW SERVICE · AI FORWARD DEPLOYED ENGINEERS
Most AI projects stall at the same point: the trial impresses, and after launch nobody uses it. A Forward Deployed Engineer (FDE) sits inside your department, watches how the work actually gets done, finds the steps that eat the most hours, and starts building that same week. Ship, measure, repeat, until the savings show up in the numbers.
ON SITE · on-site and remote, at the same table as your team
WHERE
On-site and remote, in the same meetings and the same releases as your team.
CADENCE
One usable improvement every week, with no six-month wait for a big launch.
PROOF
Every AI workflow has a before and after: hours, cost per item, cycle time.
HOW IT WORKS · 12 WEEKS
Four phases, each ending with something you can see.
Weeks 1–2
PHASE 1
Sit with the front-line team, document the process and measure the baseline: hours, error rate, waiting time.
Weeks 3–6
PHASE 2
Turn the top one or two processes into AI agents or automations, connected to CRM, ERP, documents and mailboxes, live with a small group first.
Weeks 7–10
PHASE 3
Compare hours and cost saved against the baseline, correct what needs it, then roll out to the whole team.
Weeks 11–12
PHASE 4
Documentation, the eval test set and operator training go to your team; a monthly retainer follows if you want one.
WHAT FDEs BUILD
High-volume, repetitive work with a clear right answer shows results first.
Incoming mail is classified and a reply is drafted; an agent on your team approves and sends.
Drafted from your templates and the customer’s terms, so sales edits only what matters.
Fields are read and checked against rules; only the exceptions reach a person.
Summary, action items and next step, updated on the opportunity after the call.
SOPs, policies and product material, answered with the source cited.
Numbers pulled across systems, put in the standard format and sent on time.
HOW SAVINGS ARE MEASURED
A results report every month: the pre-launch baseline against today, in terms your CFO can check.
ENGAGEMENT MODELS
Sprint on one process, or embed for the full cycle. Whether to keep an FDE on afterwards is a decision for the numbers.
ONE PROCESS
One process, from baseline to live.
FULL CYCLE
Several processes through the full cycle, then handed to your team.
ONGOING
Keep tuning after launch and take on new processes.
COMPARE
| Criterion | Traditional consultancy | Hiring your own AI engineers | EKel FDE |
|---|---|---|---|
| Who does the building | Writes reports and recommendations | Starts once the hire is made | Engineers write the code and ship it |
| Time to a first result | Usually several months | Hiring alone takes months | First process live within 6 weeks |
| Who answers for the outcome | Ends when the report is delivered | You do | Tied to before-and-after metrics |
| What stays afterwards | A slide deck | One employee, if they stay | Live systems, an eval test set and a trained team |
30 MINUTES · FDE CONSULTATION
Thirty minutes: tell us which process takes the most people-hours, and we assess whether an embedded FDE is the right fit. If it is not, we will say so.
Training, advisory, implementation and operations — four phases that follow how companies really adopt AI agents: people understand it first, then decide what to do, then build it, then keep it reliable. Start with any phase.
We teach with your real work, so people leave with a working workflow — not slides.
The top three quantified AI levers in two weeks — or a full refund if there's nothing worth doing.
Agents look things up, follow processes and call systems; we connect them safely to your CRM, ERP and documents.
Repeatable test sets track accuracy, hallucination and cost, so model updates can't silently degrade quality.
From CEO to engineer, every role needs a different kind of AI fluency. Courses are designed by role and taught with your industry and your real work.
Half a day to give leadership a shared language for AI agents: what they can do, what they can't, and where to invest first.
For boards, executive teams and business-unit heads.
Everyone turns a piece of their own repetitive work into an AI workflow that works.
For non-technical teams: sales, service, finance, admin, marketing.
Equip your engineers to build production-grade agents, not just demos.
For IT departments, in-house dev teams and Salesforce developers.
Conferences, industry associations, company learning sessions — one talk that shows a room what AI can do.
For event organisers, associations and internal knowledge sessions.
Decide what to do — and what not to. Every recommendation is tied to a measurable metric, with security and compliance assessed.
Two weeks to find your top three quantified AI levers, with ROI math and a delivery roadmap.
For companies that know AI matters but aren't sure where to start.
Set the rules, permissions and accountability before everyone starts using AI their own way.
For regulated industries such as finance and healthcare, or companies where staff already use AI tools heavily.
Someone accountable for your AI direction — without hiring a full-time AI executive.
For companies that need someone accountable for AI before a dedicated AI team makes sense.
Turn the chosen use case into an agent that is live, connected to your systems and owned by someone — start with one process, prove it, then scale.
Answer customers around the clock, tidy the pipeline, draft quotes — and hand complex cases to people.
For teams with heavy service volume, or sales reps buried in admin.
Staff find SOPs, contracts and policies in one sentence — and agents run the repetitive internal processes.
For organisations with lots of documents, slow onboarding and repetitive internal processes.
Connect agents safely to CRM, ERP and internal systems — so they don't just answer, they act.
For companies that want agents to operate systems, not just chat.
AI systems are never "done". Models update, data shifts, needs grow — we keep them accurate, reliable and cost-controlled.
Every model or prompt update passes the test set first — quality can't quietly slip.
For companies with AI in production that has to stay reliable.
Build on the first agent's success, one process and one department at a time.
For companies whose first agent has proven itself and are ready to scale.
// IN-PERSON TRAINING · CLAUDE × CODEX
EKel's own Salesforce and custom-system projects are written, reviewed, tested and deployed with Claude Code and OpenAI Codex. The training material comes from that delivery work, not from a rewrite of the vendor documentation.
01 · ANTHROPIC
From everyday work to a whole codebase
WHAT YOUR TEAM LEARNS
Web and desktop apps
Documents, analysis, slides and day-to-day work
Projects and Skills
Package a department's knowledge and methods so everyone can use them
Claude Code
Reads a whole codebase, edits code and runs tests in the terminal and the IDE
02 · OPENAI
From pairing on your machine to parallel cloud tasks
WHAT YOUR TEAM LEARNS
Codex CLI and IDE extension
Write code together on your own machine
Cloud tasks
Hand several jobs to agents in parallel
Code review
One more reviewer on every PR, and this one never tires
$ claude
> Add a discount cap check to OpportunityService and cover it with tests
Reading OpportunityService.cls and its test class
2 files changed · 3 tests added
$ sf apex run test
PASS12 tests passed
$ codex review
Review done · 1 suggestion: add a case for a null discount
$
TRAINING · CLAUDE × CODEX
They follow AI-T1 to AI-T4 and go deep on these two tools. Each can run on its own, always in person at your office; pricing depends on headcount and the number of days.
Non-technical colleagues hand their daily document, spreadsheet and slide work to Claude.
Engineers learn to direct a coding agent instead of being led by it.
Apex, LWC, Flow and metadata can go to an agent too, once the guardrails are in place.
Buy the right licences, set permissions and make usage visible before rolling out company-wide.
AGENDA · AI-C2 BOOTCAMP
This is the three-day version. On the last afternoon, everyone ships a feature in their own codebase.
How an agent thinks: context, tools and permissions
Project configuration files and a first task
Planning, breaking work down and test-driven development
MCP, subagents and parallel tasks
AI cross-review and CI
Finish a feature in your codebase and demo it
IN PERSON
Private, in-person sessions at your office
LANGUAGE
Taught in Mandarin
MATERIAL
Your real work and your codebase are the training material
IN-HOUSE TRAINING · CLAUDE × CODEX
Tell us the roles on your team, the headcount and the work you want to solve, and we will adapt the syllabus to your situation.
Move from quantified opportunity to validation and operations, using KPIs, evaluations, and human review to expand or stop.

In one line: AI Advisory builds new AI capability; Studio looks after the systems you already run. Many clients use both — Studio’s systems roadmap finds where AI fits, and AI Advisory delivers it.
AI Advisory & Operations builds new AI capability — finding AI opportunities, training teams, delivering and running agents — usually starting with a 2-week audit or a 4–8 week build, optionally followed by a monthly Fractional AI Lead. Studio looks after the systems you already run — across Salesforce, monday.com, custom apps and integrations — with a CTA on a monthly retainer. Many clients use both.
Yes. Courses are designed by role: an AI strategy briefing for leadership, hands-on workshops for non-technical teams, an agent-engineering bootcamp for developers, plus conference talks and in-house sessions. Every course is taught with your industry and your real work, so people leave with a workflow that works rather than a stack of slides.
A chatbot answers; an agent looks things up, follows a process and calls systems to get a job done. We start with one measurable use case, connect the agent safely to your CRM, ERP and internal documents, set permissions and human approval gates, and launch only after an eval test set confirms quality — then scale to other processes once it has proven itself.
A fixed-scope, two-week diagnostic. We map your operating processes, identify the top three quantified AI levers, and deliver ROI math plus a security/compliance assessment. If we find no opportunity worth doing, you get a full refund.
A senior consultant who acts as your company's AI technical lead on a part-time, monthly basis — setting AI strategy, governing adoption decisions, and standing up evaluation (eval) practices — without the cost and hiring risk of a full-time AI executive. It suits companies that need someone accountable for AI direction before a dedicated AI team is justified.
We start from specific processes, not technology in the abstract. Each AI lever is sized for the hours saved or revenue gained, the build and run cost, and the time to payback, with data and compliance risks flagged. You get board-ready math, not a tech deck.
We build an ongoing evaluation pipeline for production AI systems — a repeatable test set that monitors accuracy, hallucination, and regressions so a model or prompt update can't silently degrade quality. That is what turns a one-off AI pilot into a system you can depend on.
An FDE is an engineer embedded in your team. They watch how the work is done and measure a baseline, then turn the most time-consuming processes into AI agents or automations connected to your CRM, ERP, documents and inbox, and prove the hours and cost saved against that baseline. Start with a 6-week sprint or a full 12-week embed, then decide whether to continue on a monthly retainer.
Yes — always in person at your office; we don’t sell online courses. There are four: Claude for every team, a Claude Code × Codex bootcamp for engineers, agentic coding for Salesforce teams, and rollout & governance. All are private sessions taught in Mandarin, using your own work and codebase as the material — and an FDE can embed afterwards to help your team put it into practice.
Two weeks: top three quantified AI levers for your operation, with ROI math and compliance assessment. No opportunity found, full refund.
We use cookies
We use strictly necessary cookies to run this site, plus optional analytics cookies (Google Analytics) to understand how visitors use it. See our Cookie Policy and Privacy Policy.