AI writing of product descriptions and specifications

Products arriving from a supplier feed usually have no description. They have a name, a code and a price, and a page without text does not rank in search and does not sell, because the customer has nothing to read before deciding.

SoftCave platform executive report: business indicators, channel efficiency, sales funnel drop-off and delivery risksCOMMERCE INTELLIGENCE | EXECUTIVE EMAIL BRIEFRecipient: Owner / Managing DirectorSTRATEGIC FILTERS & BUSINESS RULESRule: High Traffic / Low ConvFunnel Leak: Checkout drop > 40%Risk: cash-on-delivery returns > 8%Monthly report (May 2026)Export PDF report1KEY BUSINESS INDICATORS (EXECUTIVE KPI SNAPSHOT)REVENUE REALISED (ERP)0 €32.400 €76.850 €114.200 €138.900 €142.500 €▲ +8.2% (vs GA4 ordered: €158k)NET CONTRIBUTION (MARGIN - ADS)0 €6.500 €15.800 €23.400 €28.200 €Gross margin: €44,745 (31.4%)BLENDED CAC / AVERAGE BASKET0.00 € / 0.00 €2.10 € / 45.00 €4.50 € / 61.20 €5.42 € / 68.50 €CAC/AOV share: 7.9% | Free delivery threshold: €80ORDER FULFILMENT RATE0.0%42.5%68.0%81.3%88.4%▼ 11.6% cancellations / 4.2% courier returnsPOAS (PROFIT ON ADS)0.00x1.20x2.15x2.85xROAS: 4.8x (Healthy cash flow)2CHANNEL EFFICIENCY AND ACQUISITION PROFITABILITYCHANNELSESSIONSCONV. %REALISEDROASPOASSTRATEGIC DECISIONPaid Search (Google)42.500 (34%)2.45%42.470 €4.8x1.62xScale budget +15%Paid Social (Meta)35.000 (28%)1.18%24.700 €2.4x0.88xReduce / switch off ad setOrganic Search (SEO)27.500 (22%)3.10%38.250 €--Strengthen SEO categoriesEmail / loyalty base13.750 (11%)5.20%23.200 €14.2x4.80xIncrease frequencyDirect / Referrals6.250 (5%)2.80%8.380 €--Track retentionOwner's view: Email brings the highest net margin at minimal CAC. Meta campaigns need their audience reviewed.3ECOMMERCE FUNNEL DROP-OFF (GA4)1. PRODUCT VIEWS125,000 sessions100%2. ADD TO CART10,500 sessions8.4%3. INITIATE CHECKOUT5,460 sessions4.3%4. PURCHASE3,047 orders (Overall conv: 2.43%)▼ Drop-off 91.6% (adds to basket: 8.4%)▼ Cart abandonment: 48.0% (delivery costs)▼ Checkout drop-off: 44.2% (mobile form)Profit lost at checkout: ~€18,500 in potential margin lost at the checkout step.4PROMOTION RANGE AND STOCK HEALTH[HIGH TRAFFIC / LOW CONV]SKU: 4892 - Tech Pro winter jacket1,420 visits | Conv: 0.42% | Stock: 84 pcs | Capital tied up: €4,200Decision: Cut the price by 10% to speed up turnover and free up cash flow.[HIDDEN GEM / HIGH MARGIN]SKU: 1042 - FastCharge wireless charger180 visits | Conv: 6.20% | Gross margin: 42% (€18 per unit)Decision: Move to position #1 in the category and increase the Google Shopping budget.[DEAD STOCK / NO SALES > 60 DAYS]14 SKUs (seasonal products)Total tied up at cost price: 12.450 € with no working capitalDecision: Create an outlet bundle and free up warehouse space.[RETURNS & MISSED DEMAND]SKU: 7711 (S3 trainers - returns 24%) | Enquiry: two-piece black412 searches with no results (estimated loss ~€3,200 in turnover)Decision: Fix the size chart for SKU 7711; add the requested range to purchasing.5CHARTS: DEVICES & COURIER RISK BY CITYDevice imbalance: share of visits vs revenue realisedVisits (%)Revenue (%)Mobile74% visits40% of revenueDesktop26% visits60% of revenue! Owner risk: 34% of revenue lost on phones because of how long checkout takes.Courier risk: rate of refused cash-on-delivery parcelsTOLERANCE THRESHOLD (8%)Belgrade2.1% (48% of shipments)Novi Sad2.8% (18% of shipments)Kragujevac4.2% (10% of shipments)Niš & South9.4% OVER THRESHOLDCash flow protection: SMS verification for cash on delivery enabled for red-zone cities (>8%).

Writing several thousand descriptions is not a job that gets planned, it is a job that gets postponed. So a bot writes the descriptions, by rules you set for it, and you keep the final say over every text you publish.

{ Integrations }

What you actually get

A description for products that have none

The bot goes through the catalogue and writes descriptions for items without text, based on the name, brand and attributes already in the system.

Filling in technical attributes

Besides the description, missing attributes are filled in as data - which means they immediately work as filters in the store as well.

Rules you set yourself

Tone, length and content of the text are defined by an instruction the bot receives. The same instruction applies across the catalogue, so the descriptions do not read as if ten people wrote them.

Sources per brand

For particular brands an official data source can be designated, so attributes are not invented where a verifiable figure exists.

Nothing is published without approval

Generated texts are reviewed and corrected before publishing, and whatever you edit yourself stays protected from later overwriting.

Common questions

Are such descriptions good for search?

They are, because the only alternative is an empty page. The text is unique to your site and not copied from the supplier's site, which is a common reason pages fail to rank.

Can the bot write an incorrect fact?

It can, just as a person copying from someone else's site can. That is why attributes are pulled from existing data wherever it exists, and texts are reviewed before publishing.

How many products can be processed?

Processing runs in the background in batches, so a catalogue of several thousand items is no obstacle. The speed depends on the limits of the service used.

Is it charged separately?

Use of the AI service is charged by the volume of text processed and runs on your account, so you see and control the cost.

Related to this

Want to know how this works in your case?

For a first estimate we need the number of items, your suppliers and the system you use today. The call takes fifteen minutes, after which you get scope, price and timeline.

Want to find out how our solution can help you?

Contact

form

Fill in the form below and request your personalised quote.