Pilot project: AI in technical writing
.sc-leistung{ –sc-pad: clamp(1rem, 3vw, 2rem); –sc-section-gap: clamp(2.5rem, 6vw, 4.5rem); –sc-blue: #1E4A8C; –sc-blue-soft: #e8eef9; –sc-line: #d8e0ec; –sc-text: #1a1a1a; –sc-muted: #5b6373; font-family: var(–font-primary, Inter, system-ui, sans-serif); color: var(–sc-text); line-height: 1.6; max-width: min(1240px, 95vw); margin: 0 auto; padding: 0 var(–sc-pad); }
.sc-leistung > * + *{ margin-top: var(–sc-section-gap); }
.sc-hero{ display: grid; grid-template-columns: 1.4fr 1fr; gap: clamp(1.5rem, 3vw, 3rem); align-items: end; padding-top: 1rem; }
@media (max-width: 768px){ .sc-hero{ grid-template-columns: 1fr; } }
.sc-hero-eyebrow{ display: inline-block; font-family: var(–font-mono, monospace); font-size: 0.85rem; text-transform: uppercase; letter-spacing: 0.1em; color: var(–sc-blue); margin: 0 0 0.75rem; }
.sc-hero h1{ font-size: clamp(1.75rem, 4vw, 3rem); line-height: 1.1; margin: 0 0 1.25rem; color: var(–sc-text); letter-spacing: -0.02em; }
.sc-hero p.sc-lead{ font-size: clamp(1.05rem, 1.8vw, 1.2rem); margin: 0; color: var(–sc-muted); }
.sc-hero-meta{ background: #fff; border: 1px solid var(–sc-line); border-radius: 14px; padding: 1.5rem; display: flex; flex-direction: column; gap: 0.75rem; font-size: 0.95rem; }
.sc-meta-row{ display: flex; justify-content: space-between; gap: 1rem; }
.sc-meta-row strong{ color: var(–sc-muted); font-weight: 500; font-size: 0.85rem; text-transform: uppercase; letter-spacing: 0.05em; }
.sc-meta-row span{ text-align: right; color: var(–sc-text); }
.sc-meta-row span.sc-meta-strong{ color: var(–sc-blue); font-weight: 600; }
.sc-cta-inline{ background: var(–sc-blue); color: #fff; text-align: center; padding: 0.85rem 1rem; border-radius: 10px; font-weight: 600; text-decoration: none; transition: opacity 0.2s; margin-top: 0.5rem; }
.sc-cta-inline:hover{ opacity: 0.9; color: #fff; }
.sc-section h2{ font-size: clamp(1.4rem, 3vw, 2rem); line-height: 1.15; margin: 0 0 1.5rem; color: var(–sc-text); }
.sc-section p{ margin: 0 0 1rem; max-width: 75ch; }
.sc-numbered-list{ counter-reset: sc-num; list-style: none; padding: 0; margin: 0; display: grid; gap: 1rem; }
.sc-numbered-list > li{ counter-increment: sc-num; padding: 1.25rem 1.5rem 1.25rem 4rem; border: 1px solid var(–sc-line); border-radius: 12px; background: #fff; position: relative; line-height: 1.55; }
.sc-numbered-list > li::before{ content: counter(sc-num, decimal-leading-zero); position: absolute; left: 1.25rem; top: 1.15rem; font-family: var(–font-mono, monospace); font-size: 0.95rem; font-weight: 700; color: var(–sc-blue); }
.sc-numbered-list > li strong{ color: var(–sc-text); display: block; margin-bottom: 0.25rem; }
.sc-deliverables{ display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 1rem; }
.sc-deliverable{ border-left: 3px solid var(–sc-blue); padding: 0.5rem 0 0.5rem 1.25rem; font-size: 0.95rem; line-height: 1.5; }
.sc-deliverable strong{ display: block; color: var(–sc-blue); margin-bottom: 0.25rem; font-weight: 700; font-size: 0.85rem; text-transform: uppercase; letter-spacing: 0.05em; }
.sc-faq{ border-top: 1px solid var(–sc-line); }
.sc-faq details{ border-bottom: 1px solid var(–sc-line); padding: 1rem 0; }
.sc-faq summary{ cursor: pointer; font-weight: 600; color: var(–sc-text); font-size: 1.05rem; line-height: 1.4; list-style: none; padding-right: 1.5rem; position: relative; }
.sc-faq summary::-webkit-details-marker{ display: none; }
.sc-faq summary::after{ content: “+”; position: absolute; right: 0; top: 0; font-size: 1.5rem; line-height: 1; color: var(–sc-blue); transition: transform 0.2s; }
.sc-faq details[open] summary::after{ content: “−”; }
.sc-faq details > p{ margin: 0.75rem 0 0; color: var(–sc-text); }
.sc-cta-block{ background: var(–sc-blue); color: #fff; border-radius: 16px; padding: clamp(2rem, 4vw, 3rem); }
.sc-cta-block h2{ color: #fff; font-size: clamp(1.4rem, 3vw, 2rem); margin: 0 0 1rem; }
.sc-cta-block p{ color: var(–sc-blue-soft); font-size: 1.05rem; margin: 0 0 1.5rem; }
.sc-cta-buttons{ display: flex; flex-wrap: wrap; gap: 0.75rem; }
.sc-cta-buttons a{ display: inline-block; padding: 0.85rem 1.5rem; border-radius: 10px; font-weight: 600; text-decoration: none; transition: opacity 0.2s; }
.sc-cta-buttons a.sc-cta-primary{ background: #fff; color: var(–sc-blue); }
.sc-cta-buttons a.sc-cta-secondary{ border: 1px solid #fff; color: #fff; }
.sc-cta-buttons a:hover{ opacity: 0.85; }
hr.sc-divider{ border: 0; border-top: 1px solid var(–sc-line); margin: 0; }
Three months, a defined scope, documented metrics
A pilot that is allowed to fail honestly. With a clear use case, defined success criteria and a solid scaling recommendation at the end. Instead of a marketing pilot.
Who needs this?
Documentation managers who want to try AI in practice without setting up an 18-month large-scale project. Prerequisites: maintained terminology, at least 3 languages, and a concrete pain point (long translation cycles, high consistency requirements, repetitive structural work).
How we proceed
- Pilot setup (weeks 1–2)Selection of the pilot area (one document type, one language direction). AI tool selection with a compliance review (GDPR, trade secrets). Definition of the success metrics before the pilot starts.
- Training (week 3)Training the technical writing team in prompt hygiene, post-editing routines, hallucination detection and compliance-relevant review steps.
- Pilot operation (weeks 4–10)Weekly reviews with documented findings. Adjustment of prompts, workflows or tool configuration based on real-world experience.
- Final evaluation (weeks 11–12)A consolidated evaluation of the metrics (creation time, review effort, complaint rate). A solid scaling recommendation with a concrete next step.
What you hold in your hand at the end
Frequently asked questions
What if the pilot fails?
Then that is an important result. You then have a solid data basis for why AI does not carry this use case – and you can argue that internally without having to defend yourself against model promises.
Which AI tools are used?
Depending on the use case and your compliance requirements: GPT-based tools, Claude, DeepL Pro, self-hosted LLMs (Llama, Mistral) or specialised tools such as SDL Trados Copilot.
Who does the operational work in the pilot?
Your technical writing team. Schübeler Consulting accompanies, trains and reviews – the AI outputs are produced and checked in your own company. That is the prerequisite for you to be able to continue on your own after the pilot.
What does later scaling cost?
It depends heavily on the result of the pilot. If the pilot goes well and the tool is cleanly integrated, the scaling costs are manageable (licences, rollout training, minor process adjustment). If larger preliminary work is needed, that is clearly named in the scaling plan.
Initial consultation — one hour, free of charge
In 60 minutes we clarify your specific situation. At the end you receive a written assessment of one A4 page that you can use to move forward internally.