I suspect that by "Riemann integration" he is referring to actually using a finite number of rectangles to approximate the integral- that is, Riemann sums.
Shreyarora, the formula for the accuracy is given in any Calculus book. For example, on page 487 of Salas, Hille, and Etgen's Calculus, the error in using n rectangles to approximate
\int_a^b f(t)dt
is given as less than
(f(b)- f(a))\frac{b- a}{n}
So, for example, if you use 10 rectangles to integrate
\int_0^1 x^2 dx
Your error would be less than (1-0)((1- 0)/10) or 1/10.
Of course, the trapezoidal method and Simpson's rule give better accuracy.